AUGUST 2015 T H E O B A L D E T A L . 1713

A Synoptic Classification of Inflow-Generating Precipitation in the ,

ALISON THEOBALD AND HAMISH MCGOWAN Climate Research Group, School of Geography, Planning and Environmental Management, University of Queensland, St Lucia, Queensland, Australia

JOHANNA SPEIRS Snowy Hydro, Ltd., Sydney, , Australia

NIK CALLOW Environmental Dynamics and Ecohydrology, School of Earth and Environment, University of Western Australia, Crawley, Western Australia, Australia

(Manuscript received 24 October 2014, in final form 23 April 2015)

ABSTRACT

Precipitation falling in the Snowy Mountains region of southeastern Australia provides fuel for hydroelectric power generation and environmental flows along major systems, as well as critical water resources for agricultural irrigation. A synoptic climatology of daily precipitation that triggers a quantifiable increase in streamflow in the headwater catchments of the Snowy Mountains region is presented for the period 1958–2012. Here, previous synoptic-meteorological studies of the region are extended by using a longer-term, year-round precipitation and reanalysis dataset combined with a novel, automated synoptic-classification technique. A three-dimensional representation of synoptic circulation is developed by effectively combining meteorological variables through the depth of the troposphere. Eleven distinct synoptic types are identified, describing key circulation features and moisture pathways that deliver precipitation to the Snowy Mountains. Synoptic types with the highest precipitation totals are commonly associated with moisture pathways originating from the northeast and northwest of Australia. These systems generate the greatest precipitation totals across the westerly and high-elevation areas of the Snowy Mountains, but precipitation is reduced in the eastern-elevation areas in the lee of the mountain ranges. In eastern regions, synoptic types with onshore transport of humid air from the Tasman Sea are the major source of precipitation. Strong seasonality in synoptic types is evident, with frontal and cutoff-low types dominating in winter and inland heat troughs prevailing in summer. Interaction between tropical and extratropical systems is evident in all seasons.

1. Introduction Mountains are one of only a few alpine regions in Aus- tralia. They form the highest part of the Great Dividing Inflows generated from precipitation falling in the Range and include Australia’s highest peak—Mount Snowy Mountains region provide vital water sources for Kosciusko—at 2228 m (Fig. 1). In contrast to younger irrigation and environmental flows in the agriculturally mountain ranges such as the European Alps, the topog- important and ecologically diverse Murray–Darling ba- raphy is not as steep, rugged, or high in elevation; instead, sin, as well as fuel for hydroelectric power generation. areas of undulating tablelands dominate the region. Located in southeastern Australia (SEA), the Snowy In response to a series of severe droughts in the region, construction began on the Snowy Mountains Hydro- Electric Scheme (‘‘Scheme’’ hereinafter) in 1949. The Corresponding author address: Alison Theobald, School of Scheme consists of a complex network of dams, hydro- Geography, Planning and Environmental Management, Level 4, Chamberlain Bldg., University of Queensland, St Lucia, Queens- electric power stations, tunnels, aqueducts, and pipelines land, 4072, Australia. that are able to divert eastward-flowing under the E-mail: [email protected] mountain range inland to the Murray and Murrumbidgee

DOI: 10.1175/JAMC-D-14-0278.1

Ó 2015 American Meteorological Society Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1714 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

FIG. 1. (a) Location of the Snowy Mountains region within Australia (red box) and the synoptic-analysis region (dashed line). (b) Snowy Mountains region showing the water catchment (solid black line) and precipitation gauges used in this study (red squares indicate gauges on the western elevations, blue diamonds indicate gauges on the high elevations, green triangles indicate gauges on the eastern slopes, and black dots indicate the gauges from the SILO dataset). Inflow gauges at and Yarrangobilly River are marked with asterisks.

Rivers. The Scheme provides on average 2360 Gl of water 2009), generating as much as 50% of the total annual per year for irrigation, underwriting AUD $3 billion of inflows to the hydroelectric catchments (Snowy Hydro agricultural production in the Murray–Darling basin. Limited 2003). Although the warmer months are gen- Furthermore, the water provides environmental flows erally drier, heavy rainfall can result from trajectories of along major rivers and offers a degree of flow regulation warm, moist air from lower latitudes and small-scale (Snowy Hydro Limited 2003; Ghassemi and White convective events that generate thunderstorms (Barry 2007; see also http://www.mdba.gov.au/about-basin/ 1992; Basist et al. 1994; Snowy Hydro Limited 2003). how-river-runs/murrumbidgee-catchment). The Scheme’s Orographic enhancement of precipitation also occurs in hydroelectric power generation meets the peak power summer as north and northwesterly winds flow perpen- demand for much of eastern Australia and currently dicular to the Snowy Mountains (Chubb et al. 2011). provides 32% of all renewable annual energy produc- The Snowy Mountains region lies toward the northern tion in Australia (http://www.snowyhydro.com.au/energy/ limits of the influence of the midlatitude westerly wind hydro/snowy-mountains-scheme). belt, where interaction between tropical and extra- The Snowy Mountains are typically one of Australia’s tropical weather systems is an important factor in the wettest regions, but the precipitation is highly variable. generation of precipitation (Wright 1989). Accordingly, Annual precipitation from days with $10 mm of this region is sensitive to changes in the annual cycle of precipitation in high-elevation regions of the Snowy the subtropical ridge (STR) and shifts in the westerly Mountains varied between 2800 and 760 mm between storm tracks (Drosdowsky 2005; Murphy and Timbal 1958 and 2012 (Fig. 2). High precipitation variability 2008; Verdon-Kidd et al. 2013; Timbal and Drosdowsky also exists on intra-annual time scales. Precipitation in 2013), as well as hydrological-cycle changes in a warm- winter and spring is heavily influenced by the prevailing ing climate that can dramatically affect seasonal pre- midlatitude westerly winds in conjunction with oro- cipitation and runoff (Beniston 2003; Viviroli et al. graphic enhancement, and these are commonly consid- 2011). In addition, the ascent mechanisms that are re- ered to be the wettest seasons (Snowy Hydro Limited sponsible for providing deep convective moisture and 2003; Pook et al. 2006; Cai and Cowan 2008; thus the potential for higher precipitation totals in the Ummenhofer et al. 2009; Chubb et al. 2011). High- Snowy Mountains vary seasonally. intensity and warm spring rains falling onto the snow- Historic snow-course data have shown significant de- pack are a major source of inflows (McGowan et al. clines in snow depth (Nicholls 2005; Hennessy et al.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1715

FIG. 2. (top) Annual precipitation across the western (plus signs), high (squares), and eastern (asterisks) elevations and (bottom) annual number of precipitation days $ 10 mm for the pe- riod 1958–2012.

2008) in association with rising alpine temperatures. Strengthening of the East Australia Current and asso- Climate-change modeling of snow depth by Hennessy ciated Tasman Sea warming due to climate change et al. (2003, 2008) and Hendrikx et al. (2013) shows a suggest increasing summer precipitation in SEA (Cai continuation of these declines. Regional-scale hydro- et al. 2005; Shi et al. 2008; Gallant et al. 2012). climatic modeling predicts rainfall and runoff to decrease Previous studies of synoptic circulation over SEA on average and to increase in extremity over coming have predominantly used manual-classification schemes, decades (Chiew et al. 2011). Meanwhile, demand for which are, by their nature, subjective and limited in the both water and energy are forecast to increase into the number of meteorological variables on which they are future (Christensen et al. 2007) resulting in increased based. The majority consider only the cool-season period. vulnerability of water resources that are already under Manual approaches are considered to be time-consuming, stress (Viviroli et al. 2011). Despite the importance of and, with the exception of Pook et al. (2014),themajority inflow-generating precipitation in the Snowy Mountains, of SEA studies have covered relatively short time periods there remains a knowledge gap regarding the long-term, of a few decades at most. Several SEA studies have fo- historic climatological behavior of the synoptic weather cused particularly on the period of extended drought that systems that deliver precipitation to the region. persisted for much of the 1990s and 2000s in an attempt to The cool-season synoptic circulation over various re- understand the significant precipitation decline that oc- gions of SEA has been well studied (e.g., Wright 1989; curred during this period (e.g., Risbey et al. 2013). Studies Pook et al. 2006; Landvogt et al. 2008; Risbey et al. 2009; confined to shorter periods may not be fully representative Gallant et al. 2012), although few studies relate specifi- of synoptic circulation over a multidecadal period. cally to the Snowy Mountains area (Colquhoun 1978; McGowan et al. (2009) identified the role that multi- Chubb et al. 2011; Fiddes et al. 2015). These studies decadal ocean–atmosphere teleconnections may play in commonly report cool-season precipitation declines. the hydroclimate of the Snowy Mountains region. Ac- From these cool-season-focused studies, it is widely re- cordingly, there is a need to extend current understanding ported that cold fronts and closed lows, including cutoff of synoptic circulation beyond the last few decades and to lows, are responsible for the majority of wintertime encompass year-round synoptic systems. precipitation. These types of weather systems can occur Previous manual classifications of precipitation sys- year-round (Landvogt et al. 2008; Wright 1989), how- tems worldwide and in Australia focus mainly on surface ever, and are also important for inflow generation out- pressure fields with limited analyses of mid- and upper- side the winter season, particularly when they interact level atmospheric properties. Between three and five with tropical systems. For instance, the highest 7-day synoptic types were identified for SEA (Wright 1989; accumulated rainfall total in the Snowy Mountains re- Pook et al. 2006; Landvogt et al. 2008; Chubb et al. 2011; gion fell between 27 February and 4 March 2012, causing Risbey et al. 2013). Studies that have applied automated widespread flooding (Bureau of Meteorology 2012). techniques have been based on a single atmospheric

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1716 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54 variable, usually mean sea level pressure (MSLP; observations used in this study (Fig. 1b). Some records in Whetton 1988; Kidson 2000; Jiang 2011; Renwick 2011), the SHL dataset begin in the 1950s, but many of the early because of difficulties encountered in combining multi- data records (before 1975) are discontinuous. Installation level data (Kidson 2000). Multivariable classifications of an automatic weather station within the study area in either classify each variable separately (e.g., Bettolli 1996 and a substantial increase in precipitation gauge et al. 2010) or use data-reduction techniques (e.g., Stahl network size after 2004 vastly improved data quality. et al. 2006; Moron et al. 2008). Self-organizing maps that Newer, heated and weighing gauges at higher elevations, are based on MSLP (Cassano and Cassano 2010)or some with wind shields, more accurately record snowfall. 500-hPa geopotential height (Newton et al. 2014a,b) Records from 56 SHL active gauges within the Snowy were employed in Canadian studies to assess catchment- Mountains have been used in this study, which covers the scale hydroclimatic variability. These approaches, as period of 1958–2012. Data are recorded instantaneously noted by Stahl et al. (2006), conceal the complex three- and were aggregated to daily totals to 0900 local time (in dimensional nature and internal variability of synoptic line with the BoM convention for daily precipitation ob- types. Consequently, causes of variability in seasonal servations). SHL data from two inflow-recording stations, rainfall that depend not only on the annual cycle of Yarrangobilly River at Ravine station and the Snowy surface pressure, but also on seasonal changes to the River above Guthega station (Fig. 1b), were also aggre- atmospheric circulation in the mid- and upper tropo- gated to daily totals. sphere (Pook et al. 2006), may not be identified. Data from 10 gauges within the Snowy Mountains Here we present a seasonal, 55-yr (1958–2012) synoptic were selected from the SILO dataset (Fig. 1b) and used climatology of inflow-generating precipitation days for to ensure a continuous daily precipitation record. The the Snowy Mountains region. A novel, objective method SILO dataset consists of observations from BoM gauges, is developed that describes synoptic types on the basis of a filled in with interpolated values where observations are suite of 21 meteorological variables throughout the depth unavailable. Rainfall is interpolated using the geo- of the troposphere. Use of reanalysis data and an auto- statistical method of ordinary kriging (Jeffrey et al. mated approach allows a multidecadal time period to be 2001). An interpolated surface (0.05830.058 grid reso- investigated and removes much of the subjectivity of lution) is produced for each day, from which missing manual classifications (Yarnal 1993). Important is that data values for point locations are extracted from the this approach allows robust conclusions to be drawn with nearest grid cell. The provision of daily data for in- regard to patterns of synoptic circulation that are re- dividual stations rather than a gridded product allows sponsible for inflow-generating precipitation, enabling direct insertion of data (Pook et al. 2010) into disconti- ongoing research to investigate trends in synoptic circu- nuities in the SHL record. For any point location, the lation variability and their significance for local to re- nearest 30 stations and all within 100 km are used for the gional hydroclimate. Such knowledge is essential to better interpolation, whichever is greater. It has been shown understand the drivers of variability in historical records that normalized precipitation removes the component of precipitation so as to make better-informed water- of precipitation variability that is due to topographic management decisions (Viviroli et al. 2011). influences and can be reliably interpolated at time scales A description of the data and methods used is outlined from hourly to monthly (Jeffrey et al. 2001). Topo- in section 2. Results of the synoptic climatology and the graphic effects are subsequently accounted for by de- variability of the synoptic weather types are presented in normalizing the interpolated surface to derive the final section 3. A discussion and conclusions are presented in rainfall surface. The SILO dataset is available for the section 4. period from 1889 to present. Precipitation data from all seasons were considered. The standard climatologi- 2. Data and methods cal seasons have been used throughout this study: December–February (DJF), March–May (MAM), a. Data June–August (JJA), and September–November (SON). A network of private, tipping- and weighing-bucket For the purpose of producing a synoptic climatology, precipitation gauges operated by Snowy Hydro Limited the European Centre for Medium-Range Weather (SHL) and the Queensland government’s Scientific In- Forecasts (ECMWF) ERA-Interim and ERA-40 re- formation for Land Owners (SILO) ‘‘Patched Point analysis products (Dee et al. 2011) were used as input to a Dataset’’ (https://www.longpaddock.qld.gov.au/silo/ppd/ clustering algorithm and to construct composite maps of index.php?reset5reset, accessed January 2014; Jeffrey the synoptic atmospheric circulation associated with et al. 2001) from the Australian Bureau of Meteorology precipitation days. ERA-Interim spans the period from (BoM) recording stations provide the daily precipitation 1979 to near–real time and is available at a default 0.758

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1717 latitude 3 0.758 longitude grid resolution. ERA-40 spans b. Methods the period 1958–2002 and has a default resolution of 2.5832.58, but a 0.75830.758 resolution can also be 1) QUALITY CONTROL obtained and is used here. Reanalysis products at For the purposes of this study, all data were subject to coarser scales have been shown in previous studies to be quality control and any data flagged as unsuitable for unsuitable for regional climate assessments (Eichler and climatological purposes were automatically removed. Gottschalck 2013) and to have difficulty accurately de- Data were checked on a gauge-by-gauge basis for tecting features such as surface pressure fronts and anomalous values by calculating the maximum, mini- troughs. As a result, past synoptic studies have needed to mum, and range. Furthermore, any data with a quality supplement coarse-resolution reanalysis data with sat- flag indicating potentially bad data were subject to ad- ellite imagery and manual-analysis charts (e.g., Pook ditional quality control and in a few cases were removed et al. 2006). Furthermore, ECMWF reanalyses have because of anomalously high half-hourly values. In ad- improved representation of Southern Hemisphere high- dition, SHL data coded as ‘‘good’’ but with a zero latitude atmospheric circulation when compared with amount where the corresponding SILO dataset showed a other reanalysis products (Marshall 2003). value greater than zero were disregarded when calculat- ERA-Interim represents the latest reanalysis product ing the daily mean values. (at the time of writing) from ECMWF and has addressed several data-assimilation issues that were encountered 2) PRECIPITATION THRESHOLD in ERA-40 (Dee et al. 2011). Comparison of clustering results from the two products for an overlapping 22-yr This study investigates days on which precipitation period (1979–2001) demonstrated no significant differ- generates inflow to reservoirs within the Snowy Moun- ence in the output (not shown). This result is in agree- tains water catchment (Fig. 1b). To define a threshold ment with Hoskins and Hodges (2005) who evaluated amount for a precipitation day, a Lyne and Hollick filter the impact of changes to observing systems in their (Lyne and Hollick 1979) was applied to separate analysis of Southern Hemisphere storm tracks. They ‘‘quickflow’’ (surface runoff) from base flow in the in- concluded that their climatological description re- flow dataset (Nathan and McMahon 1990) at the Snowy mained robust between different reanalyses and the pre- River and Yarrangobilly River stations. As alpine and and postsatellite eras. subalpine sites, respectively, these two stations provide Daily mean values of the following variables were indicative conditions across the region. Precipitation used in this study: MSLP; 500-hPa geopotential height; from the most relevant gauges to the inflow stations was 1 wind vectors at the surface, 850, 700, 500, and 250 hPa; then correlated with quickflow at a lag time of 1 day relative humidity and temperature at 850, 700, 500, and (based on prior knowledge of the behavior of pre- 250 hPa; and (the computed) 1000–500-hPa atmospheric cipitation and runoff in these catchments). The appli- thickness. The subtropical jet stream (STJ), the strength cation of the Lyne and Hollick filter to inflow data and position of which over Australia is known to in- resulted in a threshold precipitation amount of 10 mm, fluence steering and development of synoptic systems, above which quickflow and therefore inflow was en- was calculated as the magnitude of the wind vector at hanced (not shown). 250 hPa (Risbey et al. 2009). For our purposes, tem- Only precipitation during the period from December perature and humidity profiles are standard airmass in- to April was considered for the determination of this dicators of atmospheric stability and available moisture, threshold. During the cool season, precipitation can be respectively (Davis and Kalkstein 1990). Wind vectors stored as snowpack or can enhance inflow during spring provide critical information on direction of moisture snowmelt, and, particularly during this period, a static transport and steering of synoptic systems (Pook et al. precipitation–runoff threshold does not apply. It is ac- 2006). Mean sea level pressure provides an indicator of knowledged that such a precipitation–runoff threshold the large-scale current atmospheric state (Eder et al. is typically dynamic and depends upon numerous 1994) while thickness provides information on advection catchment conditions but is necessary for the purpose of and frontal position (e.g., Pook et al. 2006). Variables synoptic classification of precipitation days in this study. were obtained for the synoptic analysis area bounded by 3) GROUPING DATA latitudes 208–468S and longitudes 1208–1608E(Fig. 1a). This region is considered to be extensive enough to Different precipitation regimes are experienced capture all synoptic weather systems affecting the across the Snowy Mountains region, in part because of Snowy Mountains, including those originating in, and topographic interaction with the prevailing flow (Chubb interacting with, tropical latitudes. et al. 2011; Fiddes et al. 2015). Following Chubb et al.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1718 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

TABLE 1. Statistics of grouped precipitation gauges.

Avg Avg annual Correlation No. gauges No. gauges Group elev (m) precipitation (mm) between gauges prior to 2004 after 2004 Western 391 1246 .0.78 2 5 High 1484 1602 .0.74 8 45 Eastern 1107 857 .0.70 3 6

(2011), the 56 SHL gauges were divided into western-, lower skill scores for the eastern and western elevations, high-, and eastern-elevation groups (Fig. 1b) on the basis only 4.5%–4.9% of days are detected solely in of the similarity of the statistics shown in Table 1. Al- these groups. though the high-elevation group contains a greater A daily precipitation amount for each group was cal- number of gauges, most of these began operating after culated as the mean daily precipitation from each gauge 2004. Prior to 2004, the number of gauges was more within the group (Chubb et al. 2011; Dai et al. 2014; similar across all groups (Table 1). Such regionalization Fiddes et al. 2015). Following WMO guidelines and of data has been used successfully in previous studies Chubb et al. (2011), a minimum of four data values were in a variety of locations worldwide, helping to reduce used to calculate a mean (WMO 2011). When this was noise and spatial autocorrelation in the data (Whetton unachievable using only SHL data, the SILO data were 1988; Widmann and Schär 1997; Kidson 2000; Chubb included in the calculation. The change in the number et al. 2011; Plavcova et al. 2014). of SHL gauges, particularly within the high-elevation A continuous daily precipitation record back to 1958 group, was shown to have minimal effect on the calcula- was not possible using only the SHL data; therefore, tion of the mean, with an average difference of 0.11 mm comparisons between the SHL and SILO datasets were between mean values. Prior to the commencement of carried out to determine the suitability, and associated data recording in the western (eastern) elevations by the uncertainty, in combining the two datasets. Correla- Snowy Hydro Scheme in 1961 (1960), daily totals were tions, contingency tables, and their associated skill-score calculated solely from the SILO dataset. statistics [bias, probability of detection (POD), and Each day with a precipitation total greater than the mean absolute error (MAE)] were calculated (Wilks threshold amount of 10 mm was extracted from the daily 2006; Beesley et al. 2009; Tozer et al. 2012). Data from record. The corresponding ECMWF reanalysis vari- all 10 SILO gauges within the watershed boundary ables for each of those days were collated into a data provided the best comparison with the most recent SHL matrix and standardized on a monthly basis (over the data (which include high-resolution, heated, and fenced full study period), using z scores, to account for the in- gauges), suggesting that using all gauges in the SILO consistent units (Yarnal 1993; Hart et al. 2006; Wilson dataset back to 1958 provides the most robust estimate et al. 2013; Gao et al. 2014). Standardizing climate data of precipitation. has been demonstrated to have the additional benefit of Contingency-table statistics were better overall for removing seasonal variations in intensity of synoptic gauges within the high-elevation group, with POD systems, allowing comparison between data with dif- scores indicating that at least 78%–84% of precipitation ferent variabilities and means (Gao et al. 2014; Jiang days over 10 mm are accurately detected by SILO. The 2011; Wilks 2006; Kidson 2000; Yarnal 1993). The strong MAE between the SILO and SHL datasets varies be- influence of the seasonal cycle of the STR on weather tween 0.03 mm (high elevations) and 3.66 mm (eastern patterns in SEA means clustering applied to standard- elevations), in good agreement with Jeffrey et al. (2001) ized data better reflects the daily changes in atmospheric for the study region. SILO estimates of precipitation do circulation (Yarnal 1993). For example, Pook et al. not demonstrate a forecast bias for the high elevations, (2006, their Fig. 3) show the variability in monthly mean but there is a tendency to underforecast (overforecast) MSLP across the cool season. In our study, the monthly precipitation days after (before) 2006 in the western and standardized MSLP data reveal that precipitation days eastern elevations. Lower skill scores in the western and are always associated with negative anomalies, which eastern elevations are likely due to fewer gauges in these become stronger during the cooler months and are at a areas. Particularly in the east, the only two gauges within maximum in May (not shown). Similar month-to-month the SILO dataset are located in the southern part of the variations are apparent in the other variables used in this catchment. Lack of representation in the northern study (not shown for brevity). The standardized data catchment may explain the lower skill there. Despite the matrix was used as input into the cluster analysis.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1719

4) CLUSTER ANALYSIS of key surface and upper-air features (Davis and Kalkstein 1990; Yarnal 1993), and the presence of cloud bands. To generate synoptic types for the period 1958–2012, Classification on the basis of these variables revealed cluster analysis was applied to the standardized meteo- several days that could have been placed into more than rological variables for each daily precipitation total one cluster. Further inspection of additional reanalysis- greater than the threshold amount of 10 mm. The cluster generated variables (temperature, humidity, and wind analysis was performed across all seasons simulta- vectors) showed that most days could belong to only one neously, given the prior removal of the seasonal cycle by particular cluster. Overall, only 8% of days were moved data standardization. Equal weighting was given to each into a different cluster on the basis of the manual anal- variable, given the importance of each in influencing the ysis, from that generated by the k-means algorithm. synoptic circulation and precipitation received, as out- lined in section 2a. Hierarchical average-linkage clus- tering gave an initial indication of the number of groups 3. Results and analysis contained in the data (Wilks 2006; Hart et al. 2006; a. Synoptic classification of precipitation-bearing Trauth 2007). The k-means clustering method of Wilson systems et al. (2013), with a city-block distance measure, was then applied to the standardized variables to assign each The application of the threshold precipitation precipitation day to a synoptic type (Hart et al. 2006; amount to daily precipitation for the period of 1958– Wilson et al. 2013). The iterative nature of the k-means 2012 resulted in 3443 days being identified with inflow- technique refines the clusters by reclassifying days until generating precipitation and requiring synoptic clas- the smallest within-cluster solution is found and days sification. A day was classified if at least 10 mm of with similar meteorological characteristics are classified precipitation was recorded in the western, high, or in the same cluster (Hart et al. 2006). The algorithm was eastern group. These specific days account for almost tested for a range of clusters between 2 and 20. Exami- 40% of all precipitation days, that is, those for which nation of a plot of the distance measure against number precipitation $ 1 mm is recorded. It is acknowledged of clusters for the point at which the line flattens out, and that this definition of a precipitation day does not ac- after which distance increases again, commonly gives an count for multiday precipitation events, in which a indication of the optimum number of clusters (Wilks series of synoptic types may traverse the region as a 2006; Tan et al. 2006; Wilson et al. 2013). This was used precipitation-bearing weather system evolves through in conjunction with physical interpretation of composite time. Instead, each individual day is assigned to a maps (generated from the mean value of all days as- specific synoptic type, following Pook et al. (2006) and signed to each cluster; Kalkstein et al. 1987). The k-means Risbey et al. (2009). Figure 2 shows the annual pre- technique was initialized several times using random cipitation across the different elevations (Fig. 2a)and subsets of the data as cluster seed values. The iteration for the annual number of precipitation days of at least which the sum of distances was smallest was then used as 10 mm (Fig. 2b) between 1958 and 2012, demonstrat- the cluster seeds for the full dataset. ing the high degree of interannual variability in the Comparison of clustering results between the ERA- precipitation of the Snowy Mountains region. A sta- Interim and ERA-40 reanalyses for an overlapping 22-yr tistically significant trend in precipitation of 138 mm 2 period produced no significant differences in the (10 yr) 1 in the eastern elevations is apparent, but resulting synoptic types, with a hit rate of .80% in the western and high elevations and annual precipitation assignment of individual days to the same synoptic type. days exhibit nonsignificant decreases. This is in agreement with Hoskins and Hodges (2005) A two-sided Wilcoxon–Mann–Whitney test demon- whose comprehensive comparison of synoptic clima- strated that the median rainfall amount in correspond- tologies remained robust between different reanalyses ing clusters of the manual and automated classification products and pre- and postsatellite eras. was equal at a 95% confidence level ( p , 0.05). Fur- The automated clustering procedure was validated by thermore, the manual classification confirmed that the comparison with a manual classification for a 5-yr period automated scheme was capable of detecting expected (2008–12) and nonparametric hypothesis testing of the synoptic patterns. The types were not as clearly defined precipitation assigned to each cluster. Surface and 500-hPa as in previous manual-classification studies (e.g., Pook height charts, readily available from BoM (http:// et al. 2006; Chubb et al. 2011), however, likely because www.bom.gov.au/australia/charts/archive/), and NOAA of the greater number of variables being considered and satellite imagery (http://www.ncdc.noaa.gov/gibbs/year) thus the greater possible combinations of variables as were used to classify each day on the basis of identification well as the multitype nature of some synoptic systems.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1720 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

each of these solutions further suggested that 11 syn- optic types better represented known synoptic systems and reinforced initial hierarchical clustering results. A Wilcoxon–Mann–Whitney test determined that median precipitation between types was significantly different, rejecting the null hypothesis that all medians were equal ( p , 0.05), for 80% of all possible combinations (con- sidering precipitation across all areas of the catchment). Given that only those days experiencing precipitation $ 10 mm have been classified, rather than rain versus no- rain days, and the natural variability in precipitation amount from individual occurrences of the same type, it is more likely that the null hypothesis will not be re- jected in all cases. This result, along with physical in- terpretation of composite maps for each reanalysis FIG. 3. Within-cluster sum of distances. Similar to a scree plot, the point at which the plot flattens out (shown by the circled area) parameter (Kalkstein et al. 1987; Wilson et al. 2013), indicates that there are 11 clusters in the data. demonstrates that each of the 11 clusters represents specific synoptic types. Composite charts showing average meteorological The k-means clustering method, applied to the daily, conditions for a number of key parameters for each of standardized reanalysis data for a range of cluster the 11 synoptic types are presented in Figs. 4–9. In ad- numbers k, suggested the optimum number of clusters to dition to the clustering parameters, an analysis of co- be 10 or 11 (Fig. 3). Comparison of composite maps for lumnar precipitable water (PW) and relative vorticity

FIG. 4. Composite maps for each synoptic type showing MSLP (colored contours; hPa) and 500-hPa height (contour lines; m). Longitude and latitude are displayed on the x and y axes, respectively. The area represented in these composite maps is the synoptic-analysis region as shown in Fig. 1a.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1721

21 FIG.5.AsinFig. 4, but showing the subtropical jet stream strength at the 250-hPa level (colored contours; m s ) and direction (black arrows). for each synoptic type was conducted, and it is shown in conducive to northwest cloud bands (NWCBs). For T4, Fig. 10. These additional parameters give further in- however, NWCBs are detected on only ;10% of oc- formation on available moisture and system develop- currences, when conditions match those in Wright (1989, ment, respectively. Some between-type similarities exist their Fig. 3). NWCBs form over the warm surface waters in individual variables for a given level (e.g., MSLP), but to the northwest of Australia, where deep convection when all variables are considered together each type feeds moisture from the tropical Indian Ocean along the has distinct characteristics and distinguishing features. cloud band to southeastern Australia (Tapp and Barrell Table 2 summarizes the three-dimensional characteris- 1984; Sturman and Tapper 2006). This circulation, evi- tics, ascent mechanisms, and moisture pathways for each dent in Figs. 4 and 5 for T8 (and to a lesser extent T4 and synoptic type. The frequency of occurrence of each T11), features moisture aligned with the core of the STJ synoptic type and the resulting precipitation contribu- and its region of maximum intensity [as shown in Tapp tions in each elevation group are presented in Table 3. and Barrell (1984)]. Together these three synoptic types The synoptic classification reveals that 8 of the 11 account for 26% of all days over 10 mm. In a similar types (all except T1, T5, and T9) represent atmospheric way, and as confirmed by the manual classification, circulation patterns with a connection to tropical lati- T10-classified synoptic types have circulation that is tudes, in particular where a north or northwesterly air- conducive to cloud bands extending northward along flow (specifically between 700 and 500 hPa; Figs. 7–9) the east coast, commonly seen as an easterly dip and advects a conveyor of warm, moist air originating from cloud-band pattern (Gallant et al. 2012; Fiddes et al. the warm oceans surrounding tropical Australia toward 2015). Downstream anticyclones and ridging—apparent the Snowy Mountains (Figs. 6–8). These tropical- in T3, T4, T6, T7, T8, T10, and T11—contribute to connected systems deliver over 70% of total precipita- tropical moisture transport and enhancement of warm tion greater than 10 mm across the whole catchment air advection (WAA; Fig. 6), which, combined with an (Table 3). Three of the tropical-connected synoptic anticlockwise rotation of winds with height (‘‘backing’’), types—T8, T11, and T4—display synoptic circulation signifies forced ascent (Figs. 7–9). With the exception of

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1722 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

FIG.6.AsinFig. 4, but showing 1000–500-hPa thickness (colored contours; m) and MSLP (line contours; hPa).

T11, all tropical-connected systems demonstrate upper- extratropical fronts to the south of Australia (Sturman level divergence in the exit quadrant of the STJ (Fig. 5). and Tapper 2006; Gallant et al. 2012). Localized accel- All synoptic types exhibit airflow directions that are eration of the STJ core along the enhanced thickness conducive to orographic enhancement of precipitation. gradient, jet-stream divergence in the poleward exit Synoptic types T1, T2, T5, T6, and T8 can be grouped quadrant, and strong WAA (associated with strong as cold-cored frontal-type days [including contributions downstream anticyclones) are consistent with enhanced from closed and cutoff lows (T2 and T6) and prefrontal system development and higher precipitation totals troughs (T5 and T8)], and together account for 53% of (Risbey et al. 2009). Cold fronts (T1) occur at a fre- all classified precipitation days (Table 3). Cutoff lows quency that is similar to those of other frontal types, here follow the definition in the SEA studies of Pook although they deliver smaller amounts of precipitation et al. (2006), Risbey et al. (2009),andChubb et al. across all elevations, consistent with lower humidity and (2011), among others, in which closed circulation can be precipitable water. Strong cold-air advection (CAA), apparent at the surface or midlevels with a trough above clockwise rotation of winds with height, and jet-streak or below, respectively. The manual analysis showed that divergence downstream of the front indicate ascent is T2 fulfils the traditional criteria of cutoff lows on ap- provided primarily by frontal lift. Development of proximately 60% (50%) of occurrences and T6 fulfils closed and cutoff lows is associated with a localization them on 72% (44%) of occurrences when considering and concentration of the STJ, often forming on the MSLP (500-hPa geopotential height). In addition, poleward side of the jet. This relationship is represented closed-low types have associated fronts on 75% (T2) in Fig. 5, which shows a strong jet core located to the and 45% (T6) of occurrences. Cold-frontal types have north of the closed low for types 2 and 6. Similarly, for in common differential cyclonic vorticity advection classifications that include the passage of a cold front (CVA) at 500 hPa as an ascent mechanism (Fig. 10), but (T1 and, to some extent, T5), meridional excursions of closed-low types demonstrate stronger cyclonic vortic- the polar jet that cause it to merge with the subtropical ity maxima than do embedded cold fronts and pre- jet near the location of the front are a known feature frontal types. The types T4 and T7 show the signature of (Sturman and Tapper 2006; Risbey et al. 2009). Fur- heat lows and troughs interacting with cold-cored thermore, Risbey et al. (2009) associate cyclonic

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1723

FIG.7.AsinFig. 4, but for relative humidity (colored contours; %) and wind direction (black arrows) at 850 hPa. curvature of the jet-stream core, the exit region of high PW to the northeast and northwest of the study which is divergent, with the largest amounts of pre- region (Figs. 6 and 10), along with divergence in the cipitation (.5 mm and, in particular, for synoptic systems poleward exit quadrant of the jet stream (Fig. 5)and that produce $ 15 mm of precipitation) from frontal orographic enhancement. As a result, thermal wind systems—evident here in Fig. 6 for frontal-type days is enhanced and directed southeastward along the over 10 mm. thickness gradient, causing acceleration of the STJ Synoptic type T3 is representative of inland heat (Fig. 5)—apparent here for those synoptic systems troughs extending from a low pressure center in north- classified as T2, T6, and T8. Classifications that include ern Australia—known locally as the Cloncurry low closed lows alone account for approximately 20% of (Sturman and Tapper 2006; Gallant et al. 2012). Strong days and contribute 26%, 23%, and 16% of total pre- WAA, into the divergent region of the jet stream, is cipitation to the western, high, and eastern groups, apparent in the vicinity of the trough (Fig. 6). respectively. Topographic interaction of the prevailing airflow in The contribution of precipitation from frontal systems each synoptic type creates differences in precipitation and the midlatitude westerly airflow is reduced in east- contributions between elevation areas. Synoptic types ern elevations, in the lee of the mountain range. Instead, generating the greatest precipitation totals across onshore easterlies in the subtropics associated with westerly elevations are the result of approaching cold downstream anticyclones or ridging that advect warm, fronts or troughs, closed lows and troughs that extend humid air from a moisture corridor along the east coast toward northwest Western Australia (T2, T4, T6, T7, via inland, meridional troughs (T3, T7, and T10), and and T8; Table 3). This is similar for the high elevations, offshore lows (T9), are the major sources of precipita- where NWCBs associated with days that are classified tion (Table 3). WAA is a common ascent mechanism for as type 8 and with closed lows (T6) bring the highest these types. Differences in the spatial distribution of percentage of precipitation totals per day. A common precipitation between synoptic types shown here are feature in each of these types (T2, T4, T6, T7, and T8) consistent with investigations by Chubb et al. (2011) and is a downstream anticyclone, with WAA and relatively Fiddes et al. (2015).

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1724 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

FIG.8.AsinFig. 7, but for 700 hPa. b. Seasonality of synoptic types systems in the lower-to-midtroposphere (surface–500 hPa; Figs. 11 and 12). However, this study highlights that a sig- Clear seasonality in the frequency of synoptic types is nificant proportion of inflow-generating precipitation days evident in Fig. 11, which reflects seasonal movement of of $10 mm (approximately 20%) are recorded during the STR. The mean contribution of each type to sea- summer months, often related to the occurrence of inland sonal precipitation accumulations (Fig. 12) demon- heat troughs (i.e., T4 and T7) and increased convection. A strates the high degree of intra-annual variability. The further 20% of days occurred in the transitional autumn greatest between-type variability, in terms of both fre- season. Mean daily precipitation from all synoptic types is quency and precipitation, occurs in winter and summer. similar in all seasons, but the fewer number of precipitation Orographic enhancement of precipitation for all types is days occurring in summer and autumn and the higher PW evident, with highest elevations consistently receiving indicate a higher intensity for warm-season precipitation the largest precipitation totals in all seasons (Fig. 12). days (Table 4). The majority of types can produce seasonal precipita- In austral summer (DJF), precipitation is dominated tion accumulations beyond the 95th percentile (.2stan- by warm-cored heat-trough types T3, T4, and T7. Sum- dard deviations) and could be considered as extreme mer types are associated with weaker midlevel troughs (Pook et al. 2012). Notable large falls of precipitation, and weaker cyclonic vorticity, often displaced upstream exceeding 300 mm in the western and high elevations, have of the Snowy Mountains. Instead, stronger ridges, higher resulted from the dominant types T4 and T7 in summer moisture, enhanced WAA, and low-level (upper level) (not shown). As noted in section 3a, these types display convergence (divergence) dominate and act as ascent properties that are consistent with enhanced system de- mechanisms. Moist air, with high PW, is entrained from velopment, strong WAA, and higher precipitation totals. tropical latitudes toward the Snowy Mountains region Table 4 summarizes the mean precipitation from all (Figs. 6–8, 10). These systems deliver relatively consis- days per season and reinforces previous studies that this tent falls across all elevations (Fig. 12). region is dominated by cool-season precipitation (nearly The occurrence of each synoptic type and their asso- 60%), often associated with frontal and closed-low ciated precipitation are more consistent in autumn

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1725

FIG.9.AsinFig. 7, but for 500 hPa.

(MAM), although T1, T3, T4, and T5 dominate slightly, situated off the east coast (T9) are the dominant source representing a mix of warm- and cold-core synoptic of precipitation. types. This is a transition season with precipitation still Similar to autumn, spring (SON) observes an even generated from a moist airflow and WAA from tropical spread in the percentage of total seasonal precipitation latitudes (T3 and T4). However, the northward move- days, although cold-core types T2, T6, and T8 along with ment of the STR and midlatitude westerly wind belt is T4 occur slightly more frequently. Prefrontal and closed- apparent as prefrontal troughs and cold fronts associ- low systems (T2, T6, and T8) make the highest contri- ated with embedded lows in the Southern Ocean (T1 butions to precipitation totals, but, given that this is a and T5) begin to cross the region more frequently. transitional season, systems connected to tropical lati- Cold-cored frontal types T1, T2, T5, T6, and T8 and tudes are also common (e.g., T4 and T7). Occurrence of offshore low pressure centers (T9) dominate in winter these warm systems, on top of a late-season isothermic (JJA). All demonstrate CVA with a maximum either (‘‘ripe’’) snowpack, is considered to generate the greatest over SEA (T1, T2, and T9) or upstream (T5, T6, and T8). snowmelt and inflow in the Snowy Mountains. Winter types generally display higher maximum cy- clonic vorticity than do summer types. Only one-half of 4. Discussion and conclusions these types have a tropical moisture corridor at 700 hPa, and the influence of the midlatitude westerly wind belt is Presented here is a 55-yr (1958–2012) synoptic cli- clear, with a moisture source in the lower atmosphere matology of daily synoptic circulation systems that de- (850 hPa) over the Southern Ocean or Tasman Sea. liver greater than 10 mm of precipitation for the Snowy Low-level convergence, a northwest moisture corridor, Mountains region of southeastern Australia. This is the and WAA are common features for T6 and T8, which first study to link synoptic circulation throughout the deliver the highest winter precipitation totals. In eastern tropospheric column to precipitation at the surface in elevations the influence of orography is apparent, the Snowy Mountains. The use of a suite of variables with much lower precipitation resulting from the pre- throughout the depth of the troposphere, applied to a dominantly westerly flow. Instead, low pressure centers large gridded analysis area, expands on previous studies

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1726 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

25 21 FIG. 10. As in Fig. 4, but for columnar precipitable water (colored contours; mm) and relative vorticity at 500 hPa 310 s . Solid vorticity contours indicate positive (anticyclonic) vorticity, and dotted contours indicate negative (cyclonic) vorticity.

(e.g., Wilson et al. 2013) and results in 11 synoptic types. catchments of Australia’s iconic and economically im- The clustering method reveals subtle differences in, for portant river system: the Murray–Darling basin. example, the position and orientation of surface troughs Seasonal variations in the frequency of occurrence and the location of moisture corridors, which directly of each synoptic type highlight the variability of affect the amount of precipitation received at different atmospheric circulation affecting the Snowy Mountains sites in the Snowy Mountains region. Some types display region. Although certain synoptic systems are pre- similar attributes at certain levels, but each represents a dominant during the cool or warm seasons, they may particular synoptic atmospheric circulation for this re- occur at any time (Wright 1989), and a clear seasonal gion. The method used in this study has demonstrated signal in their incidence is apparent (Fig. 11). that difficulties in combining variables from different The seasonal movement of the STR is critical to the atmospheric levels (Kidson 2000) can be overcome, synoptic systems experienced in SEA. In winter the STR providing a vertical profile of atmospheric conditions is in its most northerly position over central Australia, during specific precipitation days. It offers an automated allowing the passage of frontal systems associated with approach to the traditional map classification (Yarnal the midlatitude westerly wind belt to traverse southern 1993) and generates synoptic types with no a priori Australia. Accordingly, winter synoptic types frequently forcing of the clustering algorithm, minimizing sub- relate to the passage of cold fronts and closed lows and jectivity. The importance of moisture source regions and are typically associated with CVA. Areas of descent ascent mechanisms in delivering precipitation to the coincide with CAA while CVA and ascending motion region of interest is demonstrated. are enhanced downwind of the front or trough (Risbey The use of a daily precipitation threshold of $10 mm et al. 2009). Jet streaks display greater intensity in win- differs from previous southeastern Australia studies in ter. Divergence in their poleward exit quadrant, along which all precipitation days have been considered. Im- with CVA in midlevels, results in rising motion (Sturman portant is that it allows classification of synoptic circu- and Tapper 2006) and is a common ascent mechanism in lation associated with precipitation days that trigger a winter. In addition, the highest mean winter precipitation quantifiable increase in streamflow in the headwater totals result from types demonstrating convergence at

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1727

TABLE 2. Description of key characteristics for each synoptic type. Directional information is referred to withstandard compass-point notation: north (N), west (W), south (S), east (E) and corresponding points between.

Synoptic type Synoptic summary Ascent mechanism Moisture pathway T1: embedded cold Cold fronts over or rapidly approaching Frontal lift; CVA; convergence of polar Southern Ocean; no fronts region; cold core; midlevel trough; no jet and STJ over Tasmania; weak tropical moisture connection to tropics downstream divergence in poleward connection at any exit quadrant of jet streak tropospheric level T2: closed lows Closed lows, including cutoff lows WAA over enhanced by Southern Ocean 850 hPa; centered over Tasmania; cold core; downstream ridge; CVA; divergence NW tropical 700 hPa weak downstream ridge; strong in poleward exit quadrant of STJ jet trough at 500 hPa, with cutoff streak over Snowy Mountains region circulation on ;50% of occasions T3: inland heat Heat troughs extend equatorward from Convergent airflow at 850 hPa; winds NE tropical 850–500 hPa; troughs low pressure center in northwest back strongly; strong WAA, enhanced high PW content Queensland; warm core; downstream by downstream ridging; weak (;40 mm) along NE anticyclone; midlevel trough divergence in poleward exit quadrant pathway of STJ T4: narrow, Narrow inland heat troughs aligned Convergent airflow at 850 hPa; WAA Southern Ocean and NE interacting, NW–SE, interacting with cold-core into divergent poleward exit tropical 850 hPa; NW inland heat low; strong downstream anticyclone; quadrant of STJ tropical 700–500 hPa; troughs upstream midlevel trough; NWCBs high (;40 mm) PW on ;10% of occasions along NE and NW pathways T5: prefrontal Prefrontal troughs and approaching cold CVA; divergence in poleward exit Southern Ocean; no troughs; fronts (west of Snowy Mountains); quadrant of jet streak, south of tropical moisture approaching cold core; upstream midlevel trough Snowy Mountains region connection cold fronts T6: upstream Approaching closed lows, including Downstream anticyclone enhances Southern Ocean 850 hPa; closed lows cutoff lows over bight; cold core; WAA into area of divergent NW 700–500 hPa; downstream anticyclone; upstream, poleward exit quadrant of jet streak; moderate PW strong midlevel trough, with cutoff CVA; winds back strongly (;25 mm) along circulation on ;45% of occasions NW pathway T7: broad, Broad inland heat troughs west of study Convergent airflow at 850 hPa; WAA NE tropical 850 hPa; interacting region, aligned N–S; interaction with enhanced by downstream anticyclone; NW tropical inland heat low pressure system SW of region; divergence in poleward exit quadrant 700–500 hPa; extensive troughs strong downstream anticyclone of STJ SW of Snowy Mountains area of high PW (;40 mm) along NE and NW pathways T8: approaching Approaching cold fronts and prefrontal WAA, enhanced by downstream Southern Ocean 850 hPa; prefrontal troughs aligned NW–SE; warm core; anticyclone; divergence in poleward NW tropical troughs and strong downstream anticyclone; exit quadrant of jet streak upstream 700–500 hPa; moderate cold fronts NWCBs on ;40% of occasions of Snowy Mountains; CVA; winds (;20 mm) PW along NWCBs back strongly NW pathway T9: offshore low Offshore low pressure centers and WAA on poleward side of cyclonic Tasman Sea; no tropical pressure troughs; includes east coast lows; cold circulation; CVA moisture connection systems core; midlevel closed circulation; no at any level; moderate interaction with tropics PW (;25 mm) over Tasman Sea T10: easterly Closed low pressure centers and troughs Convergent airflow at 850 hPa; NE tropical 850–500 hPa; dips aligned N–S, N, or NW of Snowy downstream ridging and associated high (.30 mm) PW Mountains and extending to tropics; WAA; divergence in poleward exit along NE pathway cold core; midlevel closed circulation quadrant of STJ T11: noninteracting Inland troughs; warm-core lows over Convergent airflow at 850 hPa; NW tropical inland heat central Australia; lack of upstream downstream anticyclone enhances 850–500 hPa; high PW troughs trough; NWCBs on ;35% of WAA; winds back strongly (;30 mm) along occasions NW pathway

850 hPa and trajectories of warm, moist air from tropical pressure, associated with the descending branch of the latitudes (T6 and T8). Hadley cell. Accordingly, frontal systems associated In summer, the southward movement of the STR sees with the midlatitude westerly wind belt are pushed south southern Australia under the influence of a band of high of Australia. Instead, downstream anticyclones and

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1728 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

TABLE 3. Percentage occurrence of each synoptic type across all elevation groups (first row) and percentage of total precipitation $ 10 mm received from each synoptic type across each elevation group and in total (last four rows) for the period of 1958–2012.

T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 % occurrence 10.4 11.3 8.6 9.7 10.5 9.6 10.3 10.8 7.3 6.2 5.3 % precipitation West 6.8 10.6 6.0 10.9 9.2 15.2 12.6 12.4 4.4 6.8 5.1 High 8.8 10.9 6.0 10.4 10.9 12.4 11.0 12.1 6.1 6.1 5.3 East 4.1 6.7 13.9 9.6 7.1 9.0 11.4 8.5 10.0 12.3 7.3 All elev 7.3 10.0 7.5 10.4 9.7 12.6 11.6 11.5 6.3 7.5 5.6 ridging enhance WAA and entrainment of tropical precipitation of $10 mm, confirming trajectory analyses moisture. The warmer seasons are indicative of higher conducted by McIntosh et al. (2007) and Chubb et al. precipitable water and weaker ascent (Milrad et al. (2011). In a similar way, event precipitation isotope 2014). Areas of low-level convergence that are present analyses conducted by Callow et al. (2014) identified in the dominant summer types indicate positive vertical significant synoptic-type variability in the isotopic sig- velocity and, in addition to strong WAA, provide an nature of precipitation in the Snowy Mountains. The ascent mechanism, in the absence of the CVA present in results of our study further support the dominance of winter (Risbey et al. 2009). moisture-source pathways in controlling isotopic vari- In all seasons, interaction between tropical and ability in this region. extratropical systems (Wright 1989), tropical moisture It is acknowledged that using daily variables may pathways, and vertical ascent profiles with low-level mask some of the contributions of subdaily systems such (upper level) convergence (divergence) (Gyakum as pre- and postfrontal flow to precipitation over the 2008) are highlighted as important factors generating region (Gallant et al. 2012) and excludes multiday and

FIG. 11. Intra-annual variability and relative contributions of each synoptic type to the total number of seasonal precipitation days $ 10 mm for (a) DJF, (b) MAM, (c) JJA, and (d) SON.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1729

FIG. 12. Intra-annual distribution and variability of mean precipitation (per precipitation day) associated with each synoptic type for (a) DJF, (b) MAM, (c) JJA, and (d) SON. Wider pale-gray bars represent high elevations, narrow black bars represent western elevations, and narrow dark-gray bars represent eastern elevations. multitype events. Accordingly, slow-moving synoptic used to quantify the uncertainty in using these in- systems may be overrepresented. However, when con- terpolated precipitation data. This has permitted con- sidering synoptic-scale circulation, which generally op- struction of a continuous daily precipitation record for erates on time scales of a few days, the use of daily the Snowy Mountains region; such a record is consid- variables is considered to be appropriate (Barry and ered to be essential for compiling a long-term synoptic Carleton 2001; Sturman and Tapper 2006; Pook et al. climatology (Chappell and Agnew 2001). 2006). In addition, the nature of the composite plots may In summary, this novel method proposes that synoptic result in some features in the synoptic-scale circulation typing can be successfully based on atmospheric vari- becoming masked (Milrad et al. 2014). Highly variable ables throughout the depth of the troposphere. The precipitation, complex terrain, and a relatively low higher number of types in this study compares well to density gauge network in the SILO dataset may all other automated typing schemes, which have generally contribute to interpolation errors. Extensive testing and produced more types than manual methods, with more comparison of the SHL and SILO datasets have been subtle differences between types (e.g., Newton et al.

TABLE 4. Mean number of days and seasonal precipitation received from all synoptic types across all elevations for the period of 1958–2012.

Mean No. of precipitation Percentage of total Mean seasonal Mean daily precipitation Season days $ 10 mm precipitation $ 10 mm precipitation $ 10 mm from all types (mm) DJF 12 19.6 175 43.6 MAM 13 20.8 185 43.1 JJA 20 29.6 264 39.6 SON 18 30.0 268 44.7

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1730 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

2014a,b; Plavcova et al. 2014; Kidson 2000). Although Australia and New Zealand and International Association for the k-means technique has been widely used in synoptic Mathematics and Computers in Simulation, 3886–3892. classifications, this is the first time, to our knowledge, [Available online at http://www.mssanz.org.au/modsim09/I13/ beesley.pdf.] that it has been applied to multilevel and multiparam- Beniston, M., 2003: Climatic change in mountain regions: A review eter gridded meteorological data. It has revealed the of possible impacts. Climatic Change, 59, 5–31, doi:10.1023/ complex three-dimensional nature of synoptic-scale A:1024458411589. circulation, including the importance of the influence Bettolli, M. L., O. C. Penalba, and W. M. Vargas, 2010: Synoptic of moisture pathways from tropical latitudes in the weather types in the south of South America and their re- lationship to daily rainfall in the core crop-producing region in generation of high precipitation totals. It provides a Argentina. Aust. Meteor. Oceanogr. J., 60, 37–48. method for linking regional-scale precipitation data to Bureau of Meteorology, 2012: Exceptional heavy rainfall across synoptic-scale atmospheric circulation. The spatial dis- southeast Australia. Bureau of Meteorology Special Climate tribution of precipitation associated with synoptic types Statement 39, 19 pp. [Available online at http://www.bom.gov. has implications for water resources management (Frei au/climate/current/statements/scs39.pdf.] ä Cai, W., and T. Cowan, 2008: Evidence of impacts from rising and Sch r 1998; Newton et al. 2014a,b) in this region. temperature on inflows to the Murray-Darling Basin. Geo- The synoptic-typing method that was applied here phys. Res. Lett., 35, L07701, doi:10.1029/2008GL033390. allows long-term and climatologically significant periods ——, G. Shi, T. Cowan, D. Bi, and J. Ribbe, 2005: The response of to be examined, enabling a robust investigation of the the southern annular mode, the East Australian Current, and impacts of synoptic circulation on the hydroclimate of the southern mid-latitude ocean circulation to global warming. Geophys. Res. Lett., 32, L23706, doi:10.1029/2005GL024701. the Snowy Mountains region. Future research will in- Callow, N., H. McGowan, L. Warren, and J. Speirs, 2014: Drivers of vestigate temporal variability of the synoptic types in precipitation stable isotope variability in an alpine setting, relation to interannual drivers of climate variability. Snowy Mountains, Australia. J. Geophys. Res. Atmos., 119, Increased understanding gained from this synoptic cli- 3016–3031, doi:10.1002/2013JD020710. matology has implications for water resource manage- Cassano, E. N., and J. J. Cassano, 2010: Synoptic forcing of pre- ment in regional areas, and this method could be readily cipitation in the Mackenzie and Yukon River basins. Int. J. Climatol., 30, 658–674, doi:10.1002/joc.1926. applied to other hydroclimate studies and to other re- Chappell, A., and C. T. Agnew, 2001: Geostatistical analysis and gions worldwide. numerical simulation of West African Sahel rainfall. Land Degradation, A. J. Conacher, Ed., GeoJournal Library, Vol. Acknowledgments. We thank Snowy Hydro Limited 58, Springer, 19–35. scientific staff for helpful and constructive discussions, Chiew, F. H. S., W. J. Young, and W. Cai, 2011: Current drought and and we thank three anonymous reviewers whose com- future hydroclimate projections in southeast Australia and im- ments have greatly improved this manuscript. We also plications for water resources management. Stochastic Environ. thank Joshua Soderholm for initial assistance with pro- Res. Risk Assess., 25, 601–612, doi:10.1007/s00477-010-0424-x. Christensen, J. H., and Coauthors, 2007: Regional climate pro- gramming. Alison Theobald was supported by an jections. Climate Change 2007: The Physical Science Basis, Australian Postgraduate Award and by Snowy Hydro S. Solomon et al., Eds., Cambridge University Press, 847–940. Limited. We thank Snowy Hydro Limited and the Chubb, T. H., S. T. Siems, and M. J. Manton, 2011: On the decline Queensland Government Department of Science, In- of wintertime precipitation in the Snowy Mountains of formation Technology, Innovation and the Arts for pro- southeastern Australia. J. Hydrometeor., 12, 1483–1497, doi:10.1175/JHM-D-10-05021.1. viding precipitation data. ERA-Interim and ERA-40 data Colquhoun, J. R., 1978: Snowfall on the New South Wales Snowy were provided by ECMWF (Reading, United Kingdom; Mountains. Department of Science and the Environment http://apps.ecmwf.int/datasets/). Satellite imagery was pro- Bureau of Meteorology Tech. Rep. 25, 33 pp. [Available online vided by NOAA (http://www.ncdc.noaa.gov/gibbs/year). at http://www.australianweathernews.com/snow/Colquhoun_ Snowfall_NSW_Snowy_Mountains.pdf.] Dai, J., M. Manton, S. Siems, and E. Ebert, 2014: Estimation of REFERENCES daily winter precipitation in the Snowy Mountains of south- Barry, R. G., 1992: Mountain Weather and Climate. 2nd ed. eastern Australia. J. Hydrometeor., 15, 909–920, doi:10.1175/ Routledge, 402 pp. JHM-D-13-081.1. ——, and A. M. Carleton, 2001: Synoptic and Dynamic Climatol- Davis, R. E., and L. S. Kalkstein, 1990: Development of an auto- ogy. Routledge, 620 pp. mated spatial synoptic climatological classification. Int. Basist, A., G. D. Bell, and V. Meentemeyer, 1994: Statistical re- J. Climatol., 10, 769–794, doi:10.1002/joc.3370100802. lationships between topography and precipitation patterns. Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: J. Climate, 7, 1305–1315, doi:10.1175/1520-0442(1994)007,1305: Configuration and performance of the data assimilation sys- SRBTAP.2.0.CO;2. tem. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/ Beesley, C. A., A. J. Frost, and J. Zajaczkowski, 2009: A compar- qj.828. ison of the BAWAP and SILO spatially interpolated daily Drosdowsky, W., 2005: The latitude of the subtropical ridge over rainfall datasets. Proc. 18th World IMACS/MODSIM Con- eastern Australia: The L index revisited. Int. J. Climatol., 25, gress, Cairns, Australia, Modelling and Simulation Society of 1291–1299, doi:10.1002/joc.1196.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC AUGUST 2015 T H E O B A L D E T A L . 1731

Eder, B. K., J. M. Davis, and P. Bloomfield, 1994: An automated Kalkstein, L. S., G. Tan, and J. A. Skindlov, 1987: An evaluation of classificationschemedesignedtobetter elucidate the dependence three clustering procedures for use in synoptic climatological of ozone on meteorology. J. Appl. Meteor., 33, 1182–1199, classification. J. Climate Appl. Meteor., 26, 717–730, doi:10.1175/1520-0450(1994)033,1182:AACSDT.2.0.CO;2. doi:10.1175/1520-0450(1987)026,0717:AEOTCP.2.0.CO;2. Eichler, T. P., and J. Gottschalck, 2013: A comparison of Southern Kidson, J. W., 2000: An analysis of New Zealand synoptic types Hemisphere cyclone track climatology and interannual vari- and their use in defining weather regimes. Int. J. Climatol., ability in coarse-gridded reanalysis datasets. Adv. Meteor., 20, 299–316, doi:10.1002/(SICI)1097-0088(20000315)20: 891260, doi:10.1155/2013/891260. 3,299::AID-JOC474.3.0.CO;2-B. Fiddes, S., A. B. Pezza, and V. Barras, 2015: Synoptic climatology Landvogt, P. K., J. A. T. Bye, and T. P. Lane, 2008: An in- of extreme precipitation in alpine Australia. Int. J. Climatol., vestigation of recent orographic precipitation events in 35, 172–188, doi:10.1002/joc.3970. northeast Victoria. Aust. Meteor. Mag., 57, 235–247. Frei, C., and C. Schär, 1998: A precipitation climatology of the Lyne, V., and M. Hollick, 1979: Stochastic time-variable rainfall– Alps from high-resolution rain-gauge observations. Int. J. Clima- runoff modelling. Proc. Hydrology and Water Resources tol., 18, 873–900, doi:10.1002/(SICI)1097-0088(19980630)18: Symposium, Perth, Australia, Institution of Engineers, 89–92. 8,873::AID-JOC255.3.0.CO;2-9. Marshall, G. J., 2003: Trends in the southern annular mode from Gallant, A. J. E., A. S. Kiem, D. C. Verdon-Kidd, R. C. Stone, and observations and reanalyses. J. Climate, 16, 4134–4143, D. J. Karoly, 2012: Understanding hydroclimate processes doi:10.1175/1520-0442(2003)016,4134:TITSAM.2.0.CO;2. in the Murray-Darling Basin for natural resources manage- McGowan, H. A., S. K. Marx, J. Denholm, J. Soderholm, and B. S. ment. Hydrol. Earth Syst. Sci., 16, 2049–2068, doi:10.5194/ Kamber, 2009: Reconstructing annual inflows to the headwa- hess-16-2049-2012. ter catchments of the , Australia, using the Pa- Gao, X., P. Schlosser, E. Monier, and D. Entekhabi, 2014: An cific decadal oscillation. Geophys. Res. Lett., 36, L06707, analogue approach to identify heavy precipitation events: doi:10.1029/2008GL037049. Evaluation and application to CMIP5 climate models in the McIntosh, P. C., M. J. Pook, J. S. Risbey, S. N. Lisson, and United States. J. Climate, 27, 5941–5963, doi:10.1175/ M. Rebbeck, 2007: Seasonal climate forecasts for agriculture: JCLI-D-13-00598.1. Towards better understanding and value. Field Crops Res., Ghassemi, F., and I. White, 2007: Inter-Basin Water Transfer: Case 104, 130–138, doi:10.1016/j.fcr.2007.03.019. Studies from Australia, United States, Canada, China and In- Milrad, S. M., E. H. Atallah, J. G. Gyakum, and G. Dookhie, 2014: dia. 1st ed. International Hydrology Series, Cambridge Uni- Synoptic typing and precursors of heavy warm-season pre- versity Press, 462 pp. cipitation events at Montreal, Quebec. Wea. Forecasting, 29, Gyakum, J. R., 2008: The application of Fred Sanders’ teaching to 419–444, doi:10.1175/WAF-D-13-00030.1. current research on extreme cold-season precipitation events Moron, V., A. W. Robertson, M. N. Ward, and O. Ndiaye, 2008: in the Saint Lawrence River valley region. Synoptic–Dynamic Weather types and rainfall over Senegal. Part I: Observa- Meteorology and Weather Analysis and Forecasting, Meteor. tional analysis. J. Climate, 21, 266–287, doi:10.1175/ Monogr., No. 55, Amer. Meteor. Soc., 241–250. 2007JCLI1601.1. Hart, M., R. de Dear, and R. Hyde, 2006: A synoptic climatology of Murphy, B. F., and B. Timbal, 2008: A review of recent climate tropospheric ozone episodes in Sydney, Australia. Int. variability and climate change in southeastern Australia. Int. J. Climatol., 26, 1635–1649, doi:10.1002/joc.1332. J. Climatol., 28, 859–879, doi:10.1002/joc.1627. Hendrikx, J., C. Zammit, E. Ö. Hreinsson, and S. Becken, 2013: Nathan, R. J., and T. A. McMahon, 1990: Evaluation of automated A comparative assessment of the potential impact of cli- techniques for base flow and recession analyses. Water Resour. mate change on the ski industry in New Zealand and Res., 26, 1465–1473, doi:10.1029/WR026i007p01465. Australia. Climatic Change, 119, 965–978, doi:10.1007/ Newton, B. W., T. D. Prowse, and B. R. Bonsal, 2014a: Evaluating s10584-013-0741-4. the distribution of water resources in western Canada using Hennessy, K., P. Whetton, I. Smith, J. Bathols, M. Hutchinson, and synoptic climatology and selected teleconnections. Part 1: J. Sharples, 2003: The impact of climate change on snow Winter season. Hydrol. Processes, 28, 4219–4234, doi:10.1002/ conditions in mainland Australia. CSIRO Atmospheric hyp.10233. Research Rep., 47 pp. [Available online at http://www. ——, ——, and ——, 2014b: Evaluating the distribution of water climatechange.vic.gov.au/__data/assets/pdf_file/0015/73212/ resources in western Canada using synoptic climatology and TheImpactofClimateChangeonSnowConditions2003.pdf.] selected teleconnections. Part 2: Summer season. Hydrol. ——, ——, K. Walsh, I. N. Smith, J. M. Bathols, M. Hutchinson, Processes, 28, 4235–4249, doi:10.1002/hyp.10235. and J. Sharples, 2008: Climate change effects on snow condi- Nicholls, N., 2005: Climate variability, climate change and the tions in mainland Australia and adaptation at ski resorts Australian snow season. Aust. Meteor. Mag., 54, 177–185. through snowmaking. Climate Res., 35, 255–270, doi:10.3354/ Plavcova, E., J. Kysely, and P. Stepanek, 2014: Links between cr00706. circulation types and precipitation in central Europe in the Hoskins, B. J., and K. I. Hodges, 2005: A new perspective on observed data and regional climate model simulations. Int. Southern Hemisphere storm tracks. J. Climate, 18, 4108–4129, J. Climatol., 34, 2885–2898, doi:10.1002/joc.3882. doi:10.1175/JCLI3570.1. Pook, M. J., P. C. McIntosh, and G. A. Meyers, 2006: The synoptic Jeffrey, S. J., J. O. Carter, K. B. Moodie, and A. R. Beswick, 2001: decomposition of cool-season rainfall in the southeastern Using spatial interpolation to construct a comprehensive ar- Australian cropping region. J. Appl. Meteor. Climatol., 45, chive of Australian climate data. Environ. Modell. Software, 1156–1170, doi:10.1175/JAM2394.1. 16, 309–330, doi:10.1016/S1364-8152(01)00008-1. ——, J. S. Risbey, and P. McIntosh, 2010: East coast lows, atmo- Jiang, N., 2011: A new objective procedure for classifying New spheric blocking and rainfall: A Tasmanian perspective. IOP Zealand synoptic weather types during 1958–2008. Int. Conf. Ser.: Earth Environ. Sci., 11, 012011, doi:10.1088/ J. Climatol., 31, 863–879, doi:10.1002/joc.2126. 1755-1315/11/1/012011.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC 1732 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 54

——, ——, and ——, 2012: The synoptic climatology of cool-season Tozer, C. R., A. S. Kiem, and D. C. Verdon-Kidd, 2012: On the rainfall in the Central Wheatbelt of Western Australia. Mon. uncertainties associated with using gridded rainfall data as a Wea. Rev., 140, 28–43, doi:10.1175/MWR-D-11-00048.1. proxy for observed. Hydrol. Earth Syst. Sci., 16, 1481–1499, ——, ——, and ——, 2014: A comparative synoptic climatology of doi:10.5194/hess-16-1481-2012. cool-season rainfall in major grain-growing regions of south- Trauth, M., 2007: MATLAB Recipes for Earth Sciences. 3rd ed. ern Australia. Theor. Appl. Climatol., 117, 521–533, Springer-Verlag, 336 pp. doi:10.1007/s00704-013-1021-y. Ummenhofer, C. C., A. S. Gupta, A. S. Taschetto, and M. H. Renwick, J. A., 2011: Kidson’s synoptic weather types and surface England, 2009: Modulation of Australian precipitation by climate variability over New Zealand. Wea. Climate, 31, 3–23. meridional gradients in east Indian Ocean sea surface temper- Risbey, J. S., M. J. Pook, P. C. Mcintosh, C. C. Ummenhofer, and ature. J. Climate, 22, 5597–5610, doi:10.1175/2009JCLI3021.1. G. Meyers, 2009: Characteristics and variability of synoptic Verdon-Kidd, D. C., A. S. Kiem, and R. Moran, 2013: Links be- features associated with cool season rainfall in southeastern tween the Big Dry in Australia and hemispheric multi-decadal Australia. Int. J. Climatol., 29, 1595–1613, doi:10.1002/ climate variability—Implications for water resource manage- joc.1775. ment. Hydrol. Earth Syst. Sci., 10, 13 539–13 593, doi:10.5194/ ——, ——, and ——, 2013: Spatial trends in synoptic rainfall in hessd-10-13539-2013. southern Australia. Geophys. Res. Lett., 40, 3781–3785, Viviroli, D., and Coauthors, 2011: Climate change and mountain doi:10.1002/grl.50739. water resources: Overview and recommendations for re- Shi, G., J. Ribbe, W. Cai, and T. Cowan, 2008: An interpretation of search, management and policy. Hydrol. Earth Syst. Sci., 15, Australian rainfall projections. Geophys. Res. Lett., 35, 471–504, doi:10.5194/hess-15-471-2011. L02702, doi:10.1029/2007GL032436. Whetton, P. H., 1988: A synoptic climatological analysis of rainfall Snowy Hydro Limited, 2003: Engineering Features of the Snowy variability in south-eastern Australia. J. Climatol., 8, 155–177, Mountains Scheme. 4th ed. Snowy Hydro Ltd., 20 pp. doi:10.1002/joc.3370080204. Stahl, K., R. D. Moore, and I. G. McKendry, 2006: The role of Widmann, M., and C. Schär, 1997: A principal component synoptic-scale circulation in the linkage between large-scale and long-term trend analysis of daily precipitation in ocean–atmosphere indices and winter surface climate in Switzerland. Int. J. Climatol., 17, 1333–1356, doi:10.1002/ British Columbia, Canada. Int. J. Climatol., 26, 541–560, (SICI)1097-0088(199710)17:12,1333::AID-JOC108.3.0.CO;2-Q. doi:10.1002/joc.1268. Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences. Sturman, A. P., and N. J. Tapper, 2006: The Weather and Climate of 2nd ed. Academic Press, 627 pp. Australia and New Zealand. 2nd ed. Oxford University Press, Wilson, L., M. J. Manton, and S. T. Siems, 2013: Relationship be- 541 pp. tween rainfall and weather regimes in south-eastern Queens- Tan, P., M. Steinbach, and V. Kumar, 2006: Cluster analysis: Basic land, Australia. Int. J. Climatol., 33, 979–991, doi:10.1002/ concepts and algorithms. Introduction to Data Mining, Pear- joc.3484. son Addison-Wesley, 487–568. WMO, 2011: Guide to climatological practices—2011 edition. Tapp, R. G., and S. L. Barrell, 1984: The northwest Australian WMO Publ. WMO-No. 100, 117 pp. [Available online at cloud band: Climatology, characteristics and factors associ- http://www.wmo.int/pages/prog/wcp/ccl/guide/documents/ ated with development. J. Climatol., 4, 411–424, doi:10.1002/ WMO_100_en.pdf.] joc.3370040406. Wright, W. J., 1989: A synoptic climatological classification of Timbal, B., and W. Drosdowsky, 2013: The relationship between winter precipitation in Victoria. Aust. Meteor. Mag., 37, 217– the decline of southeastern Australian rainfall and the 229. strengthening of the subtropical ridge. Int. J. Climatol., 33, Yarnal, B., 1993: Synoptic Climatology in Environmental Analysis: 1021–1034, doi:10.1002/joc.3492. A Primer. John Wiley and Sons, 256 pp.

Unauthenticated | Downloaded 09/29/21 02:26 PM UTC