A New Daily Pressure Dataset for Australia and Its Application to the Assessment of Changes in Synoptic Patterns during the Last Century

Lisa V. Alexander School of Geography and Environmental Science, Monash University, Clayton, Victoria, Australia

Search for other papers by Lisa V. Alexander in
Current site
Google Scholar
PubMed
Close
,
Petteri Uotila School of Geography and Environmental Science, Monash University, Clayton, Victoria, Australia

Search for other papers by Petteri Uotila in
Current site
Google Scholar
PubMed
Close
,
Neville Nicholls School of Geography and Environmental Science, Monash University, Clayton, Victoria, Australia

Search for other papers by Neville Nicholls in
Current site
Google Scholar
PubMed
Close
, and
Amanda Lynch School of Geography and Environmental Science, Monash University, Clayton, Victoria, Australia

Search for other papers by Amanda Lynch in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

A high-quality daily dataset of in situ mean sea level pressure was collated for Australia for the period from 1907 to 2006. This dataset was used to assess changes in daily synoptic pressure patterns over Australia in winter using the method of self-organizing maps (SOMs). Twenty patterns derived from the in situ pressure observations were mapped to patterns derived from ERA-40 data to create daily synoptic pressure fields for the past century. Changes in the frequencies of these patterns were analyzed. The patterns that have been decreasing in frequency were generally those most strongly linked to variations in the southern annular mode (SAM) index, while patterns that have increased in frequency were more strongly correlated with variations in the positive phase of El Niño–Southern Oscillation. In general, there has been a reduction in the rain-bearing systems affecting southern Australia since the beginning of the twentieth century. Over the past century, reductions in the frequencies of synoptic patterns with a marked trough to the south of the country were shown to be linked to significant reductions in severe storms in southeast Australia and decreases in rainfall at four major Australian cities: Sydney, Melbourne, Adelaide, and Perth. Of these, Perth showed the most sustained decline in both the mean and extremes of rainfall linked to changes in the large-scale weather systems affecting Australia over the past century. The results suggest a century-long decline in the frequency of low pressure systems reaching southern Australia, consistent with the southward movement of Southern Hemisphere storm tracks. While most of these trends were not significant, associated changes in rainfall and storminess appear to have had significant impacts in the region.

* Current affiliation: Climate Change Research Centre, University of New South Wales, Sydney, Australia.

Corresponding author address: Lisa Alexander, Climate Change Research Centre, University of New South Wales, Sydney, NSW 2052, Australia. Email: l.alexander@unsw.edu.au

Abstract

A high-quality daily dataset of in situ mean sea level pressure was collated for Australia for the period from 1907 to 2006. This dataset was used to assess changes in daily synoptic pressure patterns over Australia in winter using the method of self-organizing maps (SOMs). Twenty patterns derived from the in situ pressure observations were mapped to patterns derived from ERA-40 data to create daily synoptic pressure fields for the past century. Changes in the frequencies of these patterns were analyzed. The patterns that have been decreasing in frequency were generally those most strongly linked to variations in the southern annular mode (SAM) index, while patterns that have increased in frequency were more strongly correlated with variations in the positive phase of El Niño–Southern Oscillation. In general, there has been a reduction in the rain-bearing systems affecting southern Australia since the beginning of the twentieth century. Over the past century, reductions in the frequencies of synoptic patterns with a marked trough to the south of the country were shown to be linked to significant reductions in severe storms in southeast Australia and decreases in rainfall at four major Australian cities: Sydney, Melbourne, Adelaide, and Perth. Of these, Perth showed the most sustained decline in both the mean and extremes of rainfall linked to changes in the large-scale weather systems affecting Australia over the past century. The results suggest a century-long decline in the frequency of low pressure systems reaching southern Australia, consistent with the southward movement of Southern Hemisphere storm tracks. While most of these trends were not significant, associated changes in rainfall and storminess appear to have had significant impacts in the region.

* Current affiliation: Climate Change Research Centre, University of New South Wales, Sydney, Australia.

Corresponding author address: Lisa Alexander, Climate Change Research Centre, University of New South Wales, Sydney, NSW 2052, Australia. Email: l.alexander@unsw.edu.au

1. Introduction

Climate modeling studies suggest that under anthropogenic climate change storm tracks, that is, tracks of surface low pressure systems or individual storms, will shift poleward in the future (e.g., Bengtsson et al. 2006; Lynch et al. 2006), at least partly associated with shifts in the zonal sea surface temperature gradient. Studies using reanalysis data in the Southern Hemisphere suggest that this might already be occurring with indications of a southward shift in strong cyclones (Wang et al. 2006), a decrease in the frequency of low pressure systems crossing southwest Australia (Hope et al. 2006), and a decrease in cyclone system density (Pezza et al. 2007). A likely impact of this for Australia would be a reduction in the frequency of rain-bearing systems in southern regions, and there is already evidence that this is contributing to the large-scale drying in southwest Western Australia (Hope et al. 2006; IOCI 2005; Power et al. 2005; Frederiksen and Frederiksen 2007; Bates et al. 2008).

The location of storm tracks and the structure, density, and number of extratropical cyclones in the Southern Hemisphere has been well researched in recent years over the period for which reanalysis data have been available, that is, approximately the last 50 yr (e.g., Simmonds and Keay 2000; Lynch et al. 2006; Lim and Simmonds 2007; Pezza et al. 2007; Wang et al. 2006). However, most of the work on centennial and longer time scales has been confined to the Northern Hemisphere and mostly over the North Atlantic and northern Europe, where there are very long records of in situ pressure observations (WASA Group 1998; Alexandersson et al. 2000; Bärring and von Storch 2004; Ansell et al. 2006; Matulla et al. 2008; Wang et al. 2009). Studies show that while the period covered by reanalysis has shown marked variation in storm activity, it may not have been exceptional when taken in the context of a longer-term climate perspective (Allan et al. 2009; Bärring and Fortuniak 2009). An interesting question would be whether it would be possible to draw similar conclusions for the Southern Hemisphere; that is, would the recent southward shifts implied from changes in synoptic patterns defined from reanalyses data show up as unusual from a century-long perspective, for example? The lack of long-term records in the Southern Hemisphere is partly due to the much smaller amount of landmass situated under the storm track. However, at least in winter much of southern Australia is influenced by this storm track, especially from the frontal systems associated with low pressure cyclones. Recently, the Australian Bureau of Meteorology (BoM) digitized historical subdaily station and mean sea level pressure data for approximately 50 stations across Australia dating back as far as 1859. This long record now provides the opportunity to study synoptic patterns and trends in storminess in Australia for the first time on the time scales of the northern European studies.

The main aim of this study is therefore to use the BoM data to create a new, quality-controlled, and homogenized dataset of daily in situ pressure observations for Australia. We then show how this dataset could be used in combination with reanalysis data to classify seasonal synoptic patterns for Australia and determine if, or how, these patterns have changed over the past century. In addition, in order to examine how these synoptic patterns might be affecting climate variability in the region, we link the large-scale winter regimes with long-term variations in climate across southern Australia using a selection of indices including the southern annular mode (SAM) and El Niño–Southern Oscillation (ENSO). The first section of this article is concerned with the data quality control and homogeneity of the in situ pressure data, which then leads into a description of how the synoptic patterns are created. Then, long-term trends in each of the synoptic patterns are identified and are linked to changes in total and extreme rainfall and storminess over southern Australia.

2. Data quality control

Subdaily pressure data (up to eight observations a day) were digitized by the Bureau of Meteorology for 49 stations across Australia for the period prior to 1950. After 1950, pressure observations were obtained from the Bureau of Meteorology electronic database for these stations for all available observation times. Because some stations had stopped reporting in more recent decades, observations from neighboring stations were also obtained. In total, 80 stations were used in the analysis up to 2006. Figure 1 shows the locations of these stations. The earliest observations date back to 1859, but it is not until 1907 that most of the “four corners” of Australia have data, so this is the start year for this analysis. Not all data have complete records and in addition, in most cases, these data have been keyed “as read,” that is, directly transcribed from original manuscripts without quality control. Gravity and index corrections were not performed and, depending on barometer type, these could be quite large (up to several hectopascal; B. Trewin 2008, personal communication). For this reason, it was necessary to quality control and homogenize the data. The following sections describe the techniques used to ensure that the highest quality and most consistent data as possible are used in our analysis.

a. Removing erroneous values

Errors in the pressure data described here can be introduced by mistakes made when digitizing values from the original handwritten records or indeed when observers incorrectly recorded values in the original field books. In general, the majority of errors occur because digits have been transposed (Alexander et al. 2005; Wan et al. 2007). It is not always possible to identify this type of error by looking at the actual pressure values because an incorrectly recorded observation could still be a valid observation if it was within the limits of the probability distribution of pressure values. More likely these errors will be highlighted by analyzing pressure tendencies, that is, the difference between two successive pressure readings. Given that all of the stations record at local time and have differing numbers of observations during the day, data at each station were averaged to make daily mean sea level pressure (MSLP) observations relative to UTC (accounting for the introduction of daylight saving in some states in the 1970s). This meant that a comparable calculation of day-to-day pressure tendencies could be produced at all stations. MSLP tendencies were therefore calculated for each station as the difference between subsequent daily observations, and a probability distribution function (PDF) was created using all of these values. From these PDFs, the 1st and 99th percentiles were calculated and used to identify “extreme” events, that is, days with values of the tendency below the 1st percentile or above the 99th percentile. These events, while including MSLP tendencies that could correspond to actual severe storm events, should also identify the majority of digitization errors that have been produced through digit transposition. These “extreme” events were examined for each station. In the majority of cases, they highlighted individual observations either where digits had been transposed or, in several cases, where MSLP had been keyed as station level pressure or vice versa for individual or sequences of observations. If it was not clear that there was a keying error, data from neighboring stations were checked to help judge whether to amend or remove the observation.

However, this type of data quality checking cannot account for sudden nonclimatic jumps that may be introduced in the data time series through, for example, changes in observing practice or instrumentation (e.g., Wang et al. 2009). Such problems can be identified by either a thorough investigation of station metadata or by using sophisticated statistical techniques that can identify potential inconsistencies in the data. The former method is extremely time consuming, and the type of data required are rarely available (or in a convenient format). However, statistical techniques are available that can be used with or without access to original station metadata, and these techniques are described in the following section.

b. Homogeneity testing

There are many statistical techniques available to assess the consistency of climate data (e.g., Wijngaard et al. 2003; Wang 2008; Menne and Williams 2005). Here the RHTestV2 (Wang 2008) software is used because it is well tested and freely available (available online at http://cccma.seos.uvic.ca/ETCCDI/software.shtml). The method is based on the penalized maximal t (PMT) or F test (PMF) and can identify, and adjust for, multiple changepoints in a time series [see Wang (2008) and above-mentioned Web site for more details]. PMT requires the use of reference stations for the homogeneity analysis, so here we use PMF, because our stations are relatively far apart and therefore there are not always obvious neighboring stations to use for comparison. To test for potential inhomogeneities in the data, daily quality-controlled MSLP data for each station between 1907 and 2006 were fed into RHTestV2. If, for example, there is a nonclimatic jump in the data because the barometer is replaced or has changed height, then an inhomogeneity will likely exist in the data as a step change. If no dates are identified as having potential step jumps, then the station is determined to be homogeneous. This is also the case for stations with less than 20 yr of data in total because the test is less reliable for smaller amounts of data. Also, we only search for “type 1” changepoints, that is, those that are statistically significant even without access to metadata. Table 1 indicates the months and years (if any) where RHTestV2 has identified inconsistencies in the data for each of the 80 stations. In total, 21 of these stations were identified as having one or more step changes throughout the base time series. For these stations, the next step of the RHTestV2 software was run to produce an adjusted daily homogeneous series.

In some cases the step changes could indicate real climatic events. One example is ENSO, which has a major influence on the climate of Australia (McBride and Nicholls 1983). There is evidence that some of the step changes in Table 1 are in fact related to major El Niño events rather than artificial jumps in the time series. For instance, it is possible that the 1941/42 El Niño produced step changes at four of the stations. An inspection of the metadata for these stations did not suggest that there were any reasons (e.g., change in instrumentation) to indicate that these changes were anything other than real. For this reason, this time period was not corrected for biases; that is, the adjusted series discussed above is only adjusted for those changepoints that were determined to be artificial.

The quality-controlled, homogenized daily pressures produce what we believe to be the only century-long high-quality time series of daily in situ MSLP for Australia.

3. Synoptic patterns for Australia

The daily MSLP dataset described above enables us to investigate if, and how, large-scale pressure (synoptic) patterns have changed over Australia during the past century and whether any such changes are linked to changes in observed climatic events. Synoptic patterns are identified as the most typical large-scale systems that affect the weather and climate of a region. There are various well-tested “clustering” techniques that have been used to classify synoptic patterns. Philipp et al. (2006) used simulated annealing clustering to identify synoptic patterns over Europe since 1850, while Rossow et al. (2005) used K-means clustering to identify typical tropical cloud regimes from satellite data. Here a technique called self-organizing maps (SOMs) is employed, which was first applied to climate by studies such as those of Cavazos (2000), Malmgren and Winter (2000), and Hewitson and Crane (2002). Subsequently, this method has been widely used in Southern Hemisphere studies to look at large-scale drivers of climate change, for example, Lynch et al. (2006), Hope et al. (2006), and Verdon-Kidd and Kiem (2008). SOMs are now a well-tested method in climate science and have been shown to perform well when compared to other clustering algorithms (e.g., Hewitson and Crane 2002; Cassano et al. 2006). The SOM algorithm (Kohonen 2001) applies an unsupervised learning process to map input data onto the elements of a regular one- or two-dimensional array, thus providing an efficient means of interpreting and visualizing large datasets. The technique is characterized by a tendency to categorize data by preserving its probability density to produce sets of approximately equiprobable patterns or “nodes” for each SOM. The preservation of probability density is an important component of the technique because other traditional cluster analyses, such as K means, can tend to group less frequent data points in larger classes that are not necessarily representative (Michaelides et al. 2001), therefore making it difficult to pinpoint the driving mechanisms of extreme events, for example. Another advantage of SOMs is that the input data need not be spatially or temporally complete, which is usually the case when dealing with observational datasets (Samad and Harp 1992). A detailed description of the SOM technique as applied in this study is given in Cassano et al. (2006).

Although the SOM technique provides an objective method for grouping large datasets, the choice of the number of nodes to have within the SOM is somewhat subjective. Obviously, the more nodes that are defined, the smaller the aggregated Euclidean distance error from the target dataset, but this can come at the cost of having too many nodes to give a useful climate signal. Conversely, too few nodes can mean that quite different synoptic patterns may be grouped together. To optimize the SOM algorithm, SOMs using varying numbers of nodes (i.e., 6, 12, 20, 30, and 42) were calculated and assessed to find the smallest error while still maintaining meaningful synoptic climatologies. In addition, the SOM was trained by varying the following two parameters: a user-defined radius r, which is dependant on the number of nodes within the SOM and a learning rate parameter α, which was varied with each value of r (in this study, α = 0.001, 0.002, … , 0.01, 0.02, … , 0.1, 0.15, 0.2, … , 1 was tested). Each parameter decreases linearly to one (r) or zero (α) during the training of the SOM. In addition to the observed daily MSLP dataset, daily averaged fields of 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) data (Uppala et al. 2005) were also calculated between 1958 and 2001 to investigate spatially complete synoptic patterns and to compare with the in situ observations. Although studies have suggested that over the Southern Hemisphere ERA-40 is superior during the era of satellite observing compared with previous observing systems (e.g., Bengtsson et al. 2004); we found little difference between the SOMs produced between 1979 and 2001 and the longer period that we have chosen to study. The 1958–2001 period also allows for a longer overlapping period for comparison with the observed daily MSLP dataset. To allow for the seasonally varying synoptic situations that affect Australia throughout the year, the authors created SOMs separately for each season. However, only the winter [June–August (JJA)] SOM will be discussed in this paper because this is when there is more likely to be an impact from the “storm track” moving northward over southern Australia. Thus, when we refer to “the SOM” in the following sections, we are referring solely to winter patterns.

Table 2 shows the errors produced by varying the number of nodes within the SOM algorithm for both the observed daily MSLP dataset and ERA-40. The errors in the observed dataset are slightly larger for the 6- and 12-node SOMs, and this is most likely due to the greater variability of the point estimates. However, for the 20-, 30-, and 42-node SOMs the errors are almost identical between the two datasets. The optimum search radius was generally 2, but it varied up to a maximum of r = 4. From a visual inspection of the resulting patterns, it was determined that SOMs with fewer than 20 nodes did not fully represent all of the synoptic situations that occurred over Australia in winter. SOMs with more than 20 nodes, while having smaller errors, resulted in patterns that occurred infrequently, thus making further statistical analysis of the results less robust. The 20-node (4 × 5) SOM therefore was a compromise between minimizing errors and providing sufficiently different synoptic patterns to be useful for a climate change study. This is in agreement with the Hope et al. (2006) and Verdon-Kidd and Kiem (2008) studies on the synoptic influences on southwest Western Australia and Victoria, respectively, who also found that a 20-node SOM was most useful for categorizing the synoptic patterns influencing those parts of the country.

Subsequently, a single set of 20 synoptic patterns or nodes was derived by applying the SOM algorithm and using all days of daily averaged fields of MSLP between 1907 and 2006. Similarly, a 20-node SOM was produced from ERA-40 using data from 1958 to 2001 (Fig. 2). The frequency of the SOM nodes is relatively evenly distributed around 5% (with 95% confidence that the random process is in the range [4.33%, 5.67%]), with the most frequent pattern (node 13) occurring 8.2% of the time, and the least frequent pattern (node 3) occurring 3.0% of the time.

For convenience, the SOMs will be referred to as SOMOBS for those patterns derived from the in situ observations and SOMERA for those patterns derived from ERA-40. Assuming that the spatial aspects of these synoptic patterns have not changed over the last 100 yr (without making any assumptions about whether they have changed in frequency), the SOMERA patterns can be used to map to the SOMOBS patterns to produce a spatially complete picture of the changes in synoptic patterns over the last century. It might be argued that mapping stations, which are mostly located in continental Australia (see Fig. 1), to ERA-40 patterns would not be able to pick up the centers of the lows, which are generally located far to the south. However, when the domain is extended farther southward (beyond what is shown in Fig. 2), we found that the SOM patterns were dominated by the variability far to the south of Australia and were not representative of the synoptic situation over the continent. Because our interest is in capturing the low pressure systems impacting Australia, the domain shown in Fig. 2 was chosen to identify the preferred large-scale patterns over the Australian region. The mapping was done as follows.

Each station point in SOMOBS is assumed to represent the MSLP value for the nearest 1° grid box on the same grid as ERA-40. Where there was more than one station in a grid box, the value was calculated as the average of two or more stations. Given the unsupervised nature of the SOM algorithm, equivalent patterns will not necessarily appear in the same order in SOMOBS as they do in SOMERA; for example, in Fig. 2 node 1 in SOMOBS might be most closely related [in terms of giving the smallest root-mean-square error (RMSE)] to node 19 in SOMERA, node 2 in SOMOBS to node 14 in SOMERA, etc. However, if the same synoptic patterns have been driving Australian climate over time it would not be unreasonable to assume that each node in SOMOBS appears only once in SOMERA. However, finding the permutation of the 4 × 5 SOMOBS that best maps to SOMERA to produce the smallest total RMSE would require 20! (i.e., 2.4 × 1018) calculations. Therefore, another method of approximating the smallest total RMSE needs to be implemented, which does not involve performing an impossibly large number of computations.

Suppose that a represents the SOMOBS node that best maps to SOMERA node b to give the smallest error, and then the RMSE Ea,b, between the two nodes a and b, is given as
i1520-0442-23-5-1111-e1
where n represents the number of nodes in the SOM and G is the total number of grid boxes with corresponding data between SOMOBS and SOMERA. This is performed as an iterative process, such that the values of a1 and b1 in the first iteration are defined as
i1520-0442-23-5-1111-e2
In the next iteration, Eq. (1) is substituted with nodes c and d, where ca and db, such that Eq. (2) becomes (c1, d1) = (c, d), where Ec,d = min(Ec,d). This process is repeated n times (i.e., for ea, c and fb, d, etc.) until all of the nodes of SOMOBS are mapped to only one of the nodes of SOMERA. This process gives an RMSE of 44.71 hPa. To test how good this estimate is, the RMSE was calculated for 10.0 × 104 random permutations of SOMOBS and SOMERA. Errors ranged from 64.29 to 137.39 hPa. The global minimum error (i.e., the error that would be produced if a one-to-one mapping of SOMOBS to SOMERA was not assumed) is 27.26 hPa, strongly suggesting that our estimate is close to the actual minimum RMSE.

Thus, while the SOMOBS nodes calculated from 1907 to 2006 daily station MSLP are used in the analysis in the next sections, it is now possible to relate these visually to the spatially complete synoptic patterns from SOMERA (Fig. 2). In each case the SOMOBS results are presented in the order that they appear in SOMERA, making the results from Fig. 3 onward more easily comparable with the large-scale patterns in Fig. 2. Using this method, however, there is not a perfect one-to-one mapping between SOMERA and SOMOBS. For example, in the following sections the SOMOBS nodes that are mapped to SOMERA nodes 13 and 17 are actually more closely related to SOMERA node 1, which has a much more pronounced trough to the south of Australia. This needs to be taken into account when analyzing the results. However, what this method does provide is a spatially and temporally complete representation of daily synoptic patterns over Australia over the last century. For convenience, these permuted patterns of SOMOBS will be referred to in the following sections as SOM.

4. Trends in the frequency of synoptic patterns over Australia and links to changes in observed climate

Each node of the SOM can be related to the synoptic pattern on an individual day, making it possible to calculate the frequency of each large-scale pattern for all years from 1907 to 2006. It would be expected that in general the patterns toward the top (particularly the top line) of Fig. 2 would reflect the weather systems bringing rain to southern Australia, while the high pressure patterns toward the bottom (particularly the bottom right-hand corner) of the figure would be more likely to be associated with clear-sky conditions over much of southern Australia. Decadal trends calculated over the last century for each synoptic pattern are also shown in Fig. 2. In general, patterns with a marked trough to the south of Australia show a decline in frequency over the last 100 yr. The largest decline of 3 days century−1 is seen in node 13. However, this trend is not statistically significant. The only statistically significant result is a downward trend of 0.25 days decade−1 in node 17, and this could occur by chance given that there are 20 patterns. Note from the previous sections that these two nodes are actually more closely related (in terms of smallest Euclidean distance) to node 1 of the SOM. In total, eight nodes show a decline in frequency while 12 nodes show an increase. The largest increase is 0.23 days decade−1 in node 15. This node has a much less pronounced trough to the south of Australia as do nodes 14, 16, 19, and 20, which also indicate increasing trends over the last century.

Figure 3 shows the frequency of the SOM nodes in more detail. Rather than just a linear increase or decline in frequency, results indicate marked decadal variability in most of the nodes, and in some cases there is evidence of possible step changes in the time series, for example, around the mid-1960s in node 13. This may be indicative of a poleward shift of the Southern Hemisphere storm tracks as suggested by many previous studies (e.g., Hope et al. 2006; Pezza et al. 2007; Hendon et al. 2007; Wang et al. 2006). Hope et al. (2006) showed that for southwest Western Australia, the rainfall decline in winter over the last 50 yr is likely to be linked to a southward shift of rain-bearing synoptic patterns. Other studies using station data have shown mixed patterns of daily rainfall decline depending on region and season (e.g., Hennessy et al. 1999; Haylock and Nicholls 2000; Alexander et al. 2007; Gallant et al. 2007). High-quality rainfall measurements are available for station locations across Australia since 1910 (Haylock and Nicholls 2000), along with other proxy variables such as a 150-yr severe storm record at Cape Otway (Alexander and Power 2009); thus, the link between these and possible changes in synoptic patterns could be analyzed. In the next sections we analyze the link between the changing frequencies of the synoptic patterns identified here for Australia (Figs. 2 and 3) and other large-scale climate indices, such as the SAM (e.g., Hendon et al. 2007; Li et al. 2005; Cai et al. 2005; Karoly 2003; Cai et al. 2003) and ENSO (e.g., Nicholls et al. 1996; Power et al. 1998; Jones and Trewin 2000), which are known to be related to Australian climate variability. In addition the effect that these changes may have had on precipitation and storminess records will also be investigated. Significance is tested at the 5% level throughout.

a. SAM and El Niño indices

The SAM index used here for comparison is the normalized seasonal (JJA) Jones et al. (2009) station-based reconstruction available for the period from 1905 to 2005. To assess if our results are related to ENSO variability, we use the Niño-4 index, which is calculated from seasonally averaged SST anomalies in the tropical Pacific between 6°N–6°S and 160°E–150°W. The index used was calculated from the JJA SST anomalies from the 1° latitude × 1° longitude Hadley Centre Global Sea Ice and Sea Surface Temperature (HadISST1) dataset (Rayner et al. 2003) over the period from 1907 to 2006.

Figure 3 shows the correlations between each of the SOM nodes with the SAM and Niño-4 indices for the time periods for which they were available. The SAM index is significantly anticorrelated with nodes 13 and 17 in winter, with correlations of −0.58 and −0.47 (explaining 34% and 22% of the variance, respectively). From Fig. 2 we can see that there have been strong declines in these nodes over the past century, with node 17 exhibiting a statistically significant decline. In general, the patterns in the left-hand column of the SOM (apart from node 9) exhibit anticorrelations with the SAM index. This implies that decreases in the frequency of these nodes are most likely linked to an increasingly positive SAM index, assuming that this correlation remains constant through time. Such a change would reflect the poleward contraction of midlatitude westerlies (Hendon et al. 2007), leading to a southward-shifted storm track.

The Niño-4 index is significantly anticorrelated (−0.31) with node 1 (which has decreased over the last century; see Fig. 2), which is indicative of a strong relationship with the frequency of this node and the La Niña phase of ENSO. On the other hand, the frequency of node 19 (which has increased in the last century; see Fig. 2), with high pressures over continental Australia and small pressure gradients to the south, is significantly positively correlated (0.36) with the Niño-4 index, which is indicative of a strong relationship with the El Niño phase of ENSO. The frequencies of the patterns in the upper two rows of the SOM nearly all exhibit negative correlations with the Niño-4 index, while in general the bottom row patterns, and particularly the bottom right of the SOM, exhibit positive correlations.

The results do seem to indicate that the synoptic patterns calculated in this study are linked to other large-scale drivers of the Australian climate, giving us confidence in the results. Given that, in the following sections we will use the daily synoptic patterns to investigate their influence on storminess and rainfall indices across southern Australia.

b. Severe storm index

Alexander and Power (2009) describe the calculation of a severe storm index for Cape Otway on the southern coast of the state of Victoria (see Fig. 1 for the location). They show that between 1865 and 2006, there has been a significant decline in the number of severe storms at this location. This again could suggest that there has been a shift in the storm tracks related to a reduction of the essential rain-bearing weather systems reaching southern Australia. To investigate this, the large-scale pressure patterns on each date in winter between 1907 and 2006 when a severe storm occurred at Cape Otway (a total of 136 events) were analyzed to determine if any particular node or nodes were responsible for driving these events.

Figure 4a shows the percentage of severe storm events that are driven by each SOM node along with the node’s relative frequency within the SOM. Figure 4a indicates that node 13 dominates during severe storm events at Cape Otway, accounting for over 32% of events. The next most frequent driving patterns are node 9 (14% of events) and node 19 (9% of events). Because node 13 is the most frequent synoptic pattern (see Fig. 2), we also show the ratio of the percentage of each pattern relative to the frequency of that pattern within the SOM in Fig. 4b. Nodes 13, 9, and 19 still stand out clearly as the dominating patterns occurring during severe storms at Cape Otway with also a small but above-expected influence from nodes 1 and 12. The fact that node 13 stands out is perhaps to be expected because this is the node with the highest correlation with the SAM index (Fig. 3); interestingly, nodes 9 and 19 seem to be more highly correlated with ENSO. This would imply some interacting mechanisms between ENSO and SAM and their effect on severe storms at Cape Otway in winter. However, the reduction in the frequency of certain large-scale systems over Australia, particularly node 13, is certainly a reasonable explanation as to why severe storm events have significantly decreased at Cape Otway over this period.

c. Daily rainfall

To investigate how changes in the frequency of synoptic patterns may influence the amount and intensity of rainfall, daily rainfall observations from four stations located close to large urban centers in southern Australia (Sydney, Melbourne, Adelaide, and Perth1) were obtained from the high quality, post-1910 rainfall dataset described in Haylock and Nicholls (2000). Each of these cities has over 1 million inhabitants (over 10 million in total) making up over half of Australia’s population, therefore making it extremely important to understand if changes in large-scale processes are affecting the amount of rainfall reaching these areas. First, we will analyze how each of the synoptic patterns affects total rainfall (PRCPTOT) at each city, followed by an analysis of the following two measures that are defined to examine how daily rainfall events may be changing: 1) a measure of the intensity of daily rainfall (SDII) and 2) the maximum 1-day rainfall amount (RX1day) associated with each SOM node (see Alexander et al. 2007).

All days with rain (i.e., ≥1 mm) at each of the four locations were assigned a synoptic type as defined in the SOM. To determine how closely linked the synoptic types were with rainfall at each station, we first assessed how well correlated the frequencies of the large-scale patterns (Fig. 3) were to the total winter precipitation at each station. Figure 5 shows the time series of winter rainfall at Perth during each of the synoptic nodes from Fig. 2. Perth has seen a significant decline in rainfall in recent decades, and 12 of the nodes studied here indicate a decline in the total rainfall associated with them in the past century. Five of those trends are statistically significant (nodes 4, 6, 8, 13, and 17). The results from the other cities are not as straightforward (not shown). While Sydney also sees a decline at 12 nodes, only two of these are statistically significant, and Melbourne and Adelaide only see declines at seven nodes each, and only one of these at Melbourne is significant. In fact, Adelaide has had a statistically significant increase in winter rainfall at four nodes (5, 7, 12, and 15) in the last 100 yr, but in general these patterns do not bring as much rain to the region as some other patterns, so this increase might have limited impact in terms of total rainfall at Adelaide.

To see how the changing frequency of synoptic patterns over the last century might affect daily rainfall amounts recorded at each city, average daily rainfall intensities were calculated as the total rainfall each winter divided by the number of rain days (SDII) for each SOM node at each location. The authors analyzed these daily rainfall intensities for each SOM node from Fig. 2, which generally showed that (at least for Melbourne, Adelaide, and Perth) the largest intensities both in terms of means and extremes were associated with the SOM nodes from the top and left-hand side of Fig. 2 (not shown), that is, those patterns with a marked trough to the south of Australia. To put these results in context, the daily intensity of rainfall was collated for each SOM pattern and the intensities for the SOM nodes that were decreasing in frequency were compared with those that were increasing in frequency for each city between 1910 and 2006, when high-quality rainfall data were available. Figure 6 shows a box plot of the daily averaged rainfall intensity associated with both the increasing and decreasing synoptic types in Fig. 2. The boxes represent the interquartile range, the line through the boxes represents the median value, and the lower and upper “whiskers” represent the 5th and 95th percentiles, respectively, of the daily rainfall intensity distribution for the increasing and decreasing SOM nodes. From this figure it is clear that the nodes that have decreased in frequency over the past century bring much more daily rainfall to each of the cities studied both in terms of means and extremes (the exception is Melbourne, where the mean and extremes are almost identical in the two cases). This would suggest that there would be a much greater impact on the region from a decrease in the frequency of these synoptic patterns. However, the results for Melbourne could imply that changes in frontal systems might not be driving rainfall variability and this would seem to agree with other studies that suggest that in parts of Victoria it is cutoff lows that dictate the interannual variability of rainfall (e.g., Pook et al. 2006, 2009; Risbey et al. 2009).

Given that SOM node 13 appears to be important for both rainfall and storminess indicators, we further analyze some of the effects in both means and extremes of rainfall of changes in this node for each of our southern Australian cities. Figure 7 shows the century-long influence of this pattern on total winter rainfall amount (PRCPTOT), maximum 1-day winter rainfall amount (RX1day), and average intensity of daily winter rainfall (SDII) at each of the southern Australian cities studied. In each case, PRCPTOT has decreased in all four cities, and statistically significantly so in Sydney and Perth, with declines of 0.36 and 4.2 mm decade−1, respectively. Sydney and Perth also have seen statistically significant declines in RX1day (0.5 and 0.9 mm decade−1, respectively) and SDII (0.1 and 0.3 mm day−1 decade−1, respectively), but interestingly Melbourne and Adelaide have seen increases in these measures, though they were nonsignificant. It is clear that for all three rainfall measures, the results are most pronounced for Perth. Southwest Western Australia has perhaps seen the most sustained decline in average rainfall of any Australian region in the last few decades, and Hope et al. (2006) suggest that this is related to a decrease in troughs affecting the region and an increase in high pressure systems. Li et al. (2005) who looked at changes in extreme rainfall events in southwest Western Australia suggested that there was evidence for a drying of winter daily rainfall extremes after 1965 related to changes in the SAM index; however, this was only evident from one of the five stations that they studied. Thus, it would appear that while there is evidence that changes in the frequencies of synoptic patterns over the last century are strongly linked to long-term changes in rainfall across southern Australia, these changes are complex and affect the rainfall in each city differently.

5. Summary and conclusions

For the first time, a quality-controlled and homogenized daily station MSLP dataset has been produced for Australia for the period of 1907–2006. This dataset was used in conjunction with reanalysis data to produce daily synoptic patterns for Australia for the past 100 yr using self-organizing maps (SOMs). Taking winter (June–August) as an example, 20 patterns or “nodes” were identified to be the major large-scale systems affecting the climate of Australia. The frequencies of each of the synoptic patterns were analyzed over the past century. In general, there has been a decline in those patterns that have a marked trough to the south of the country while patterns with marked high pressures across continental Australia have been generally increasing in frequency. Correlations showed that over the past century, the observed variability in the SAM index and Niño-4 variations associated with La Niña were most closely linked to these former patterns, while variations associated with El Niño were most closely linked to the latter patterns.

Using a Cape Otway storm index it was shown that observed decreases in severe storm events in southeast Australia over the past century have been associated with decreases in the frequency of one particular synoptic pattern (node 13) driving these events, although two other nodes (19 and 9) also had an effect on changes in storminess. Node 13 was shown to have strong links with variations in the SAM index, while nodes 19 and 9 were more highly correlated with variations in the Niño-4 index linked with its El Niño phase.

The SOM nodes were also shown to be highly correlated with total winter rainfall amounts at Sydney, Melbourne, Adelaide, and Perth. Node 13 again showed up as producing a decline in total winter rainfall at all four cities over the last 100 yr, which was statistically significant at Sydney and Perth. Perth showed the largest declines in rainfall: 5 of the 20 nodes analyzed indicated statistically significant decreases. Daily rainfall changes were also considered. SOM nodes that have decreased in frequency over the last century are responsible for much greater daily rainfall intensity (both in terms of means and extremes) than those nodes that have increased in frequency, particularly at Sydney, Adelaide, and Perth. The largest declines were again seen in Perth where maximum 1-day rainfall amounts and daily rainfall intensity have decreased significantly. Over the last century, average daily winter rainfall intensity at Perth linked with node 13 has decreased by over 3 mm.

The results here mostly confirm the conclusions from previous studies using a similar technique over shorter time scales, namely, Hope et al. (2006) over southwest Australia and Verdon-Kidd and Kiem (2008) over southeast Australia, who showed that there had been significant changes to the synoptic systems driving rainfall in those regions. In this paper, however, we have used a new independent dataset to analyze patterns for the whole of Australia for a longer time period and we have shown that decreases in the frequency of these large-scale synoptic systems are likely to be part of a longer-term decline. The results from this study do appear to provide further independent evidence for a large-scale shift of weather patterns across Australia, linked to a decline in severe storms and rainfall across southern Australia over the past century. Northern Hemisphere studies have not shown unusual variation in more recent decades storm activity when taken in a long-term context, and the frequency of decline of the individual synoptic patterns described here is also perhaps not outside what might be expected by chance. However, the associated impacts of these long-term declines on rainfall and storminess are likely to have been significant over the past century.

Acknowledgments

This research, and NN, were partly supported by the Australian Research Council through Discovery Project DP0877417. We are grateful to Blair Trewin of the Bureau of Meteorology for providing the raw MSLP data and to the three anonymous reviewers who helped to significantly improve this manuscript.

REFERENCES

  • Alexander, L. V., and S. Power, 2009: Severe storms inferred from 150 years of sub-daily pressure observations along Victoria’s “Shipwreck Coast”. Aust. Meteor. Oceanogr. J., 58 , 129133.

    • Search Google Scholar
    • Export Citation
  • Alexander, L. V., S. F. B. Tett, and T. Jonsson, 2005: Recent observed changes in severe storms over the United Kingdom and Iceland. Geophys. Res. Lett., 32 , L13704. doi:10.1029/2005GL022371.

    • Search Google Scholar
    • Export Citation
  • Alexander, L. V., P. Hope, D. Collins, B. Trewin, A. Lynch, and N. Nicholls, 2007: Trends in Australia’s climate means and extremes: A global context. Aust. Meteor. Mag., 56 , 118.

    • Search Google Scholar
    • Export Citation
  • Alexandersson, H., H. Tuomenvirta, T. Schmith, and K. Iden, 2000: Trends of storms in NW Europe derived from an updated pressure data set. Climate Res., 14 , 7173.

    • Search Google Scholar
    • Export Citation
  • Allan, R., S. Tett, and L. Alexander, 2009: Fluctuations in autumn-winter severe storms over the British Isles: 1920 to present. Int. J. Climatol., 29 , 357371. doi:10.1002/joc.1765.

    • Search Google Scholar
    • Export Citation
  • Ansell, T. J., and Coauthors, 2006: Daily mean sea level pressure reconstructions for the European–North Atlantic region for the period 1850-2003. J. Climate, 19 , 27172742.

    • Search Google Scholar
    • Export Citation
  • Bärring, L., and H. von Storch, 2004: Scandinavian storminess since about 1800. Geophys. Res. Lett., 31 , L20202. doi:10.1029/2004GL020441.

    • Search Google Scholar
    • Export Citation
  • Bärring, L., and K. Fortuniak, 2009: Multi-indices analysis of southern Scandinavian storminess 1780–2005 and links to inter-decadal variations in the NW Europe–North Sea region. Int. J. Climatol., 29 , 373384. doi:10.1002/joc.1842.

    • Search Google Scholar
    • Export Citation
  • Bates, B. C., P. Hope, B. Ryan, I. Smith, and S. Charles, 2008: Key findings from the Indian Ocean Climate Initiative and their impact on policy development in Australia. Climatic Change, 89 , 339354. doi:10.1007/s10584-007-9390-9.

    • Search Google Scholar
    • Export Citation
  • Bengtsson, L., K. I. Hodges, and S. Hagemann, 2004: Sensitivity of the ERA40 reanalysis to the observing system: Determination of the global atmospheric circulation from reduced observations. Tellus, 56A , 456471.

    • Search Google Scholar
    • Export Citation
  • Bengtsson, L., K. I. Hodges, and E. Roekner, 2006: Storm tracks and climate change. J. Climate, 19 , 35183543.

  • 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.

    • Search Google Scholar
    • Export Citation
  • Cai, W. J., G. Shi, and Y. Li, 2005: Multidecadal fluctuations of winter rainfall over southwest Western Australia simulated in the CSIRO Mark 3 coupled model. Geophys. Res. Lett., 32 , L12701. doi:10.1029/2005GL022712.

    • Search Google Scholar
    • Export Citation
  • Cassano, J. J., P. Uotila, and A. H. Lynch, 2006: Changes in synoptic weather patterns in the polar regions in the 20th and 21st centuries. Part 1: Arctic. Int. J. Climatol., 26 , 10271049.

    • Search Google Scholar
    • Export Citation
  • Cavazos, T., 2000: Using self-organizing maps to investigate extreme climate events: An application to wintertime precipitation in the Balkans. J. Climate, 13 , 17181732.

    • Search Google Scholar
    • Export Citation
  • Frederiksen, J. S., and C. S. Frederiksen, 2007: Interdecadal changes in Southern Hemisphere winter storm track modes. Tellus, 59A , 599617.

    • Search Google Scholar
    • Export Citation
  • Gallant, A. J. E., K. J. Hennessy, and J. Risbey, 2007: Trends in rainfall indices for six Australian regions: 1910-2005. Aust. Meteor. Mag., 56 , 223239.

    • Search Google Scholar
    • Export Citation
  • Haylock, M., and N. Nicholls, 2000: Trends in extreme rainfall indices for an updated high quality data set for Australia, 1910-1998. Int. J. Climatol., 20 , 15331541.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., D. W. J. Thompson, and M. C. Wheeler, 2007: Australian rainfall and surface temperature variations associated with the Southern Hemisphere annular mode. J. Climate, 20 , 24522467.

    • Search Google Scholar
    • Export Citation
  • Hennessy, K. J., R. Suppiah, and C. M. Page, 1999: Australian rainfall changes, 1910-1995. Aust. Meteor. Mag., 48 , 113.

  • Hewitson, B. C., and R. G. Crane, 2002: Self-organizing maps: Applications to synoptic climatology. Climate Res., 22 , 1326.

  • Hope, P. K., W. Drosdowsky, and N. Nicholls, 2006: Shifts in the synoptic systems influencing southwest Western Australia. Climate Dyn., 26 , 751764.

    • Search Google Scholar
    • Export Citation
  • Indian Ocean Climate Initiative, 2005: IOCI reports key findings of recent research in south-western climate. IOCI Bulletin 6, 4 pp. [Available online at http://www.ioci.org.au/pdf/IOCI_Bulletin6.pdf].

    • Search Google Scholar
    • Export Citation
  • Jones, D. A., and B. C. Trewin, 2000: On the relationships between the El Niño-Southern Oscillation and Australian land surface temperature. Int. J. Climatol., 20 , 697719.

    • Search Google Scholar
    • Export Citation
  • Jones, J. M., R. L. Fogt, M. Widmann, G. J. Marshall, P. D. Jones, and M. Visbeck, 2009: Historical SAM variability. Part I: Century-length seasonal reconstructions. J. Climate, 22 , 53195345.

    • Search Google Scholar
    • Export Citation
  • Karoly, D. J., 2003: Ozone and climate change. Science, 302 , 236237.

  • Kohonen, T., 2001: Self-organizing maps of massive databases. Eng. Intell. Syst. Electr. Eng. Commun., 9 , 179185.

  • Li, Y., W. Cai, and E. P. Campbell, 2005: Statistical modeling of extreme rainfall in southwest Western Australia. J. Climate, 18 , 852863.

    • Search Google Scholar
    • Export Citation
  • Lim, E. P., and I. Simmonds, 2007: Southern Hemisphere winter extratropical cyclone characteristics and vertical organization observed with ERA-40 data in 1979–2001. J. Climate, 20 , 26752690.

    • Search Google Scholar
    • Export Citation
  • Lynch, A., P. Uotila, and J. J. Cassano, 2006: Changes in synoptic weather patterns in the polar regions in the 20th and 21st centuries, Part 2: Antarctic. Int. J. Climatol., 26 , 11811199.

    • Search Google Scholar
    • Export Citation
  • Malmgren, B. A., and A. Winter, 2000: Climate zonation in Puerto Rico based on principal components analysis and an artificial neural network. J. Climate, 12 , 977985.

    • Search Google Scholar
    • Export Citation
  • Matulla, C., W. Schöner, H. Alexandersson, H. von Storch, and X. L. Wang, 2008: European storminess: Late 19th century to present. Climate Dyn., 31 , 125130. doi:10.1007/s00382-007-0333-y.

    • Search Google Scholar
    • Export Citation
  • McBride, J., and N. Nicholls, 1983: Seasonal relationships between Australian rainfall and the Southern Oscillation. Mon. Wea. Rev., 111 , 19982004.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., and C. N. Williams, 2005: Detection of undocumented change points: On the use of multiple test statistics and composite reference series. J. Climate, 18 , 42714286.

    • Search Google Scholar
    • Export Citation
  • Michaelides, S., C. S. Pattichis, and G. Kleovoulou, 2001: Classification of rainfall variability by using artificial neural networks. Int. J. Climatol., 21 , 14011414.

    • Search Google Scholar
    • Export Citation
  • Nicholls, N., B. Lavery, C. Frederiksen, W. Drosdowsky, and S. Torok, 1996: Recent apparent changes in relationships between the El Niño-Southern oscillation and Australian rainfall and temperature. Geophys. Res. Lett., 23 , 33573360.

    • Search Google Scholar
    • Export Citation
  • Pezza, A. B., I. Simmonds, and J. A. Renwick, 2007: Southern Hemisphere cyclones and anticyclones: Recent trends and links with decadal variability in the Pacific Ocean. Int. J. Climatol., 27 , 14031419.

    • Search Google Scholar
    • Export Citation
  • Philipp, A., P. M. Della-Marta, J. Jacobeit, D. R. Fereday, P. D. Jones, A. Moberg, and H. Wanner, 2006: Long term variability of daily North Atlantic–European pressure patterns since 1850 classified by simulated annealing clustering. J. Climate, 20 , 40654095.

    • Search Google Scholar
    • Export Citation
  • Pook, M. J., P. C. McIntosh, and G. A. Meyers, 2006: The synoptic decomposition of cool-season rainfall in the southeastern Australian cropping region. J. Appl. Meteor. Climatol., 45 , 11561170.

    • Search Google Scholar
    • Export Citation
  • Pook, M., S. Lisson, J. Risbey, C. C. Ummenhofer, P. McIntosh, and M. Rebbeck, 2009: The autumn break for cropping in southeast Australia: Trends, synoptic influences and impacts on wheat yield. Int. J. Climatol., 29 , 20122026. doi:10.1002/joc.1833.

    • Search Google Scholar
    • Export Citation
  • Power, S., F. Tseitkin, S. Torok, B. Lavery, R. Dahni, and B. McAvaney, 1998: Australian temperature, Australian rainfall and the Southern Oscillation, 1910-1992: Coherent variability and recent changes. Aust. Meteor. Mag., 47 , 85101.

    • Search Google Scholar
    • Export Citation
  • Power, S., B. Sadler, and N. Nicholls, 2005: The influence of climate science on water management in WA: Lessons for climate scientists. Bull. Amer. Meteor. Soc., 86 , 839844.

    • Search Google Scholar
    • Export Citation
  • Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108 , 4407. doi:10.1029/2002JD002670.

    • Search Google Scholar
    • Export Citation
  • Risbey, J. S., M. J. Pook, P. C. McIntosh, C. C. Ummenhofer, and G. Meyers, 2009: Characteristics and variability of synoptic features associated with cool season rainfall in southeastern Australia. Int. J. Climatol., 29 , 15951613. doi:10.1002/joc.1775.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., G. Tselioudis, A. Polak, and C. Jakob, 2005: Tropical climate described as a distribution of weather states indicated by distinct mesoscale clod property mixtures. Geophys. Res. Lett., 32 , L21812. doi:10.1029/2005GL024584.

    • Search Google Scholar
    • Export Citation
  • Samad, T., and S. A. Harp, 1992: Self-organization with partial data. Network-Comput. Neural Syst., 3 , 205212.

  • Simmonds, I., and K. Keay, 2000: Mean Southern Hemisphere extratropical cyclone behavior in the 40-year NCEP–NCAR reanalysis. J. Climate, 13 , 873885.

    • Search Google Scholar
    • Export Citation
  • Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131 , 29613012.

  • Verdon-Kidd, D., and A. S. Kiem, 2008: On the relationship between large-scale climate modes and regional synoptic patterns that drive Victorian rainfall. Hydrol. Earth Syst. Sci. Discuss., 5 , 27912815.

    • Search Google Scholar
    • Export Citation
  • Wan, H., X. L. Wang, and V. R. Swail, 2007: A quality assurance system for Canadian hourly pressure data. J. Appl. Meteor. Climatol., 46 , 18041817.

    • Search Google Scholar
    • Export Citation
  • Wang, X. L. L., 2008: Accounting for autocorrelation in detecting mean shifts in climate data series using the penalized maximal t or F test. J. Appl. Meteor. Climatol., 47 , 24232444.

    • Search Google Scholar
    • Export Citation
  • Wang, X. L., V. R. Swail, and F. W. Zwiers, 2006: Climatology and changes of extra-tropical cyclone activity: Comparison of ERA-40 with NCEP–NCAR reanalysis for 1958-2001. J. Climate, 19 , 31453166.

    • Search Google Scholar
    • Export Citation
  • Wang, X. L., F. W. Zwiers, V. R. Swail, and Y. Feng, 2009: Trends and variability of storminess in the northeast Atlantic region, 1874-2007. Climate Dyn., 33 , 11791195. doi:10.1007/s00382-008-0504-5.

    • Search Google Scholar
    • Export Citation
  • WASA Group, 1998: Changing waves and storms in the northeast Atlantic? Bull. Amer. Meteor. Soc., 79 , 741760.

  • Wijngaard, J. B., A. M. G. Klein Tank, and G. P. Können, 2003: Homogeneity of 20th century European daily temperature and precipitation series. Int. J. Climatol., 23 , 679692.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Location of stations (⋄) used in this study. Some metropolitan centers (x) are also marked for reference. Macquarie Island (54.5°S, 158.94°E) and Heard Island (53.02°S, 73.39°E) were also digitized but are not marked on the map.

Citation: Journal of Climate 23, 5; 10.1175/2009JCLI2972.1

Fig. 2.
Fig. 2.

Winter (JJA) synoptic patterns for Australia derived from ERA-40 data between 1958 and 2001 (SOMERA). The numbers in brackets indicate the relative frequency of the associated SOMOBS nodes between 1907 and 2006 (see text for details). Decadal trends m in the SOMOBS node frequency are also shown. An asterisk after the number indicates that the trend is statistically significant at the 5% level using a nonparametric Mann–Kendall test.

Citation: Journal of Climate 23, 5; 10.1175/2009JCLI2972.1

Fig. 3.
Fig. 3.

Frequency of each SOM node from Fig. 2 from 1907 to 2006. The thick gray line indicates decadal variations in the data using a 21-term binomial filter. Correlations calculated over the period from 1907 to 2005 are also shown between the frequencies of each node and the JJA SAM index (Jones et al. 2009) and the seasonally averaged JJA Niño-4 index calculated from HadISST1 (Rayner et al. 2003).

Citation: Journal of Climate 23, 5; 10.1175/2009JCLI2972.1

Fig. 4.
Fig. 4.

(a) Bars represent the percentage of synoptic patterns occurring during winter “severe storms” at Cape Otway overlaid with the relative frequency of each node within the SOM; (b) the ratio of when synoptic patterns are driving winter severe storms at Cape Otway to the relative frequency of that pattern within the SOM is represented. The x axis represents the SOM nodes from Fig. 2.

Citation: Journal of Climate 23, 5; 10.1175/2009JCLI2972.1

Fig. 5.
Fig. 5.

Time series of total rainfall for Perth associated with each of the SOM nodes from Fig. 2. Decadal trends m are also given (mm). The line of best fit using an ordinary least squares regression (solid and dashed lines) and that these trends are significant at the 5% level using a Mann–Kendall test is indicated (solid lines).

Citation: Journal of Climate 23, 5; 10.1175/2009JCLI2972.1

Fig. 6.
Fig. 6.

The first (second) box plot in each panel represents winter (JJA) daily rainfall intensity from rain days (≥1 mm) during SOM nodes that are decreasing (increasing) in frequency. Data used from stations in the high-quality dataset (Haylock and Nicholls 2000) closest to (a) Sydney, (b) Melbourne, (c) Adelaide, and (d) Perth using all daily data from 1910 to 2005.

Citation: Journal of Climate 23, 5; 10.1175/2009JCLI2972.1

Fig. 7.
Fig. 7.

Rainfall totals (PRCPTOT), maximum daily rainfall (RX1day), and daily rainfall intensity (SDII) associated with node 13 from Fig. 2 at four Australian cities. Decadal trends m are also given (mm, PRCPTOT and RX1day; and mm day−1, SDII). The line of best fit using an ordinary least squares regression (straight solid and dashed line) and that these trends are significant at the 5% level using a Mann-Kendall test (solid lines) are indicated.

Citation: Journal of Climate 23, 5; 10.1175/2009JCLI2972.1

Table 1.

Homogeneity information on the stations used in this study. Only stations where break points have been identified and the time series adjusted are included.

Table 1.
Table 2.

Information on the minimum errors produced (with associated value of the alpha parameter) in the winter (JJA) SOM analysis using both the observed daily MSLP dataset created in this study and daily averaged MSLP ERA-40 data. Errors are calculated as the sum of all the root-mean square Euclidean distances between the SOM and the target dataset.

Table 2.

1

Stations used were Cataract Dam (Sydney), Yan Yean (Melbourne), Happy Valley Reservoir (Adelaide), and Wilgarrup (Perth), at 34.27°S, 150.81°E; 35.57°S, 145.11°E; 35.06°S, 138.56°E; and 34.15°S, 116.02°E, respectively.

Save
  • Alexander, L. V., and S. Power, 2009: Severe storms inferred from 150 years of sub-daily pressure observations along Victoria’s “Shipwreck Coast”. Aust. Meteor. Oceanogr. J., 58 , 129133.

    • Search Google Scholar
    • Export Citation
  • Alexander, L. V., S. F. B. Tett, and T. Jonsson, 2005: Recent observed changes in severe storms over the United Kingdom and Iceland. Geophys. Res. Lett., 32 , L13704. doi:10.1029/2005GL022371.

    • Search Google Scholar
    • Export Citation
  • Alexander, L. V., P. Hope, D. Collins, B. Trewin, A. Lynch, and N. Nicholls, 2007: Trends in Australia’s climate means and extremes: A global context. Aust. Meteor. Mag., 56 , 118.

    • Search Google Scholar
    • Export Citation
  • Alexandersson, H., H. Tuomenvirta, T. Schmith, and K. Iden, 2000: Trends of storms in NW Europe derived from an updated pressure data set. Climate Res., 14 , 7173.

    • Search Google Scholar
    • Export Citation
  • Allan, R., S. Tett, and L. Alexander, 2009: Fluctuations in autumn-winter severe storms over the British Isles: 1920 to present. Int. J. Climatol., 29 , 357371. doi:10.1002/joc.1765.

    • Search Google Scholar
    • Export Citation
  • Ansell, T. J., and Coauthors, 2006: Daily mean sea level pressure reconstructions for the European–North Atlantic region for the period 1850-2003. J. Climate, 19 , 27172742.

    • Search Google Scholar
    • Export Citation
  • Bärring, L., and H. von Storch, 2004: Scandinavian storminess since about 1800. Geophys. Res. Lett., 31 , L20202. doi:10.1029/2004GL020441.

    • Search Google Scholar
    • Export Citation
  • Bärring, L., and K. Fortuniak, 2009: Multi-indices analysis of southern Scandinavian storminess 1780–2005 and links to inter-decadal variations in the NW Europe–North Sea region. Int. J. Climatol., 29 , 373384. doi:10.1002/joc.1842.

    • Search Google Scholar
    • Export Citation
  • Bates, B. C., P. Hope, B. Ryan, I. Smith, and S. Charles, 2008: Key findings from the Indian Ocean Climate Initiative and their impact on policy development in Australia. Climatic Change, 89 , 339354. doi:10.1007/s10584-007-9390-9.

    • Search Google Scholar
    • Export Citation
  • Bengtsson, L., K. I. Hodges, and S. Hagemann, 2004: Sensitivity of the ERA40 reanalysis to the observing system: Determination of the global atmospheric circulation from reduced observations. Tellus, 56A , 456471.

    • Search Google Scholar
    • Export Citation
  • Bengtsson, L., K. I. Hodges, and E. Roekner, 2006: Storm tracks and climate change. J. Climate, 19 , 35183543.

  • 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.

    • Search Google Scholar
    • Export Citation
  • Cai, W. J., G. Shi, and Y. Li, 2005: Multidecadal fluctuations of winter rainfall over southwest Western Australia simulated in the CSIRO Mark 3 coupled model. Geophys. Res. Lett., 32 , L12701. doi:10.1029/2005GL022712.

    • Search Google Scholar
    • Export Citation
  • Cassano, J. J., P. Uotila, and A. H. Lynch, 2006: Changes in synoptic weather patterns in the polar regions in the 20th and 21st centuries. Part 1: Arctic. Int. J. Climatol., 26 , 10271049.

    • Search Google Scholar
    • Export Citation
  • Cavazos, T., 2000: Using self-organizing maps to investigate extreme climate events: An application to wintertime precipitation in the Balkans. J. Climate, 13 , 17181732.

    • Search Google Scholar
    • Export Citation
  • Frederiksen, J. S., and C. S. Frederiksen, 2007: Interdecadal changes in Southern Hemisphere winter storm track modes. Tellus, 59A , 599617.

    • Search Google Scholar
    • Export Citation
  • Gallant, A. J. E., K. J. Hennessy, and J. Risbey, 2007: Trends in rainfall indices for six Australian regions: 1910-2005. Aust. Meteor. Mag., 56 , 223239.

    • Search Google Scholar
    • Export Citation
  • Haylock, M., and N. Nicholls, 2000: Trends in extreme rainfall indices for an updated high quality data set for Australia, 1910-1998. Int. J. Climatol., 20 , 15331541.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., D. W. J. Thompson, and M. C. Wheeler, 2007: Australian rainfall and surface temperature variations associated with the Southern Hemisphere annular mode. J. Climate, 20 , 24522467.

    • Search Google Scholar
    • Export Citation
  • Hennessy, K. J., R. Suppiah, and C. M. Page, 1999: Australian rainfall changes, 1910-1995. Aust. Meteor. Mag., 48 , 113.

  • Hewitson, B. C., and R. G. Crane, 2002: Self-organizing maps: Applications to synoptic climatology. Climate Res., 22 , 1326.

  • Hope, P. K., W. Drosdowsky, and N. Nicholls, 2006: Shifts in the synoptic systems influencing southwest Western Australia. Climate Dyn., 26 , 751764.

    • Search Google Scholar
    • Export Citation
  • Indian Ocean Climate Initiative, 2005: IOCI reports key findings of recent research in south-western climate. IOCI Bulletin 6, 4 pp. [Available online at http://www.ioci.org.au/pdf/IOCI_Bulletin6.pdf].

    • Search Google Scholar
    • Export Citation
  • Jones, D. A., and B. C. Trewin, 2000: On the relationships between the El Niño-Southern Oscillation and Australian land surface temperature. Int. J. Climatol., 20 , 697719.

    • Search Google Scholar
    • Export Citation
  • Jones, J. M., R. L. Fogt, M. Widmann, G. J. Marshall, P. D. Jones, and M. Visbeck, 2009: Historical SAM variability. Part I: Century-length seasonal reconstructions. J. Climate, 22 , 53195345.

    • Search Google Scholar
    • Export Citation
  • Karoly, D. J., 2003: Ozone and climate change. Science, 302 , 236237.

  • Kohonen, T., 2001: Self-organizing maps of massive databases. Eng. Intell. Syst. Electr. Eng. Commun., 9 , 179185.

  • Li, Y., W. Cai, and E. P. Campbell, 2005: Statistical modeling of extreme rainfall in southwest Western Australia. J. Climate, 18 , 852863.

    • Search Google Scholar
    • Export Citation
  • Lim, E. P., and I. Simmonds, 2007: Southern Hemisphere winter extratropical cyclone characteristics and vertical organization observed with ERA-40 data in 1979–2001. J. Climate, 20 , 26752690.

    • Search Google Scholar
    • Export Citation
  • Lynch, A., P. Uotila, and J. J. Cassano, 2006: Changes in synoptic weather patterns in the polar regions in the 20th and 21st centuries, Part 2: Antarctic. Int. J. Climatol., 26 , 11811199.

    • Search Google Scholar
    • Export Citation
  • Malmgren, B. A., and A. Winter, 2000: Climate zonation in Puerto Rico based on principal components analysis and an artificial neural network. J. Climate, 12 , 977985.

    • Search Google Scholar
    • Export Citation
  • Matulla, C., W. Schöner, H. Alexandersson, H. von Storch, and X. L. Wang, 2008: European storminess: Late 19th century to present. Climate Dyn., 31 , 125130. doi:10.1007/s00382-007-0333-y.

    • Search Google Scholar
    • Export Citation
  • McBride, J., and N. Nicholls, 1983: Seasonal relationships between Australian rainfall and the Southern Oscillation. Mon. Wea. Rev., 111 , 19982004.

    • Search Google Scholar
    • Export Citation
  • Menne, M. J., and C. N. Williams, 2005: Detection of undocumented change points: On the use of multiple test statistics and composite reference series. J. Climate, 18 , 42714286.

    • Search Google Scholar
    • Export Citation
  • Michaelides, S., C. S. Pattichis, and G. Kleovoulou, 2001: Classification of rainfall variability by using artificial neural networks. Int. J. Climatol., 21 , 14011414.

    • Search Google Scholar
    • Export Citation
  • Nicholls, N., B. Lavery, C. Frederiksen, W. Drosdowsky, and S. Torok, 1996: Recent apparent changes in relationships between the El Niño-Southern oscillation and Australian rainfall and temperature. Geophys. Res. Lett., 23 , 33573360.

    • Search Google Scholar
    • Export Citation
  • Pezza, A. B., I. Simmonds, and J. A. Renwick, 2007: Southern Hemisphere cyclones and anticyclones: Recent trends and links with decadal variability in the Pacific Ocean. Int. J. Climatol., 27 , 14031419.

    • Search Google Scholar
    • Export Citation
  • Philipp, A., P. M. Della-Marta, J. Jacobeit, D. R. Fereday, P. D. Jones, A. Moberg, and H. Wanner, 2006: Long term variability of daily North Atlantic–European pressure patterns since 1850 classified by simulated annealing clustering. J. Climate, 20 , 40654095.

    • Search Google Scholar
    • Export Citation
  • Pook, M. J., P. C. McIntosh, and G. A. Meyers, 2006: The synoptic decomposition of cool-season rainfall in the southeastern Australian cropping region. J. Appl. Meteor. Climatol., 45 , 11561170.

    • Search Google Scholar
    • Export Citation
  • Pook, M., S. Lisson, J. Risbey, C. C. Ummenhofer, P. McIntosh, and M. Rebbeck, 2009: The autumn break for cropping in southeast Australia: Trends, synoptic influences and impacts on wheat yield. Int. J. Climatol., 29 , 20122026. doi:10.1002/joc.1833.

    • Search Google Scholar
    • Export Citation
  • Power, S., F. Tseitkin, S. Torok, B. Lavery, R. Dahni, and B. McAvaney, 1998: Australian temperature, Australian rainfall and the Southern Oscillation, 1910-1992: Coherent variability and recent changes. Aust. Meteor. Mag., 47 , 85101.

    • Search Google Scholar
    • Export Citation
  • Power, S., B. Sadler, and N. Nicholls, 2005: The influence of climate science on water management in WA: Lessons for climate scientists. Bull. Amer. Meteor. Soc., 86 , 839844.

    • Search Google Scholar
    • Export Citation
  • Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108 , 4407. doi:10.1029/2002JD002670.

    • Search Google Scholar
    • Export Citation
  • Risbey, J. S., M. J. Pook, P. C. McIntosh, C. C. Ummenhofer, and G. Meyers, 2009: Characteristics and variability of synoptic features associated with cool season rainfall in southeastern Australia. Int. J. Climatol., 29 , 15951613. doi:10.1002/joc.1775.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., G. Tselioudis, A. Polak, and C. Jakob, 2005: Tropical climate described as a distribution of weather states indicated by distinct mesoscale clod property mixtures. Geophys. Res. Lett., 32 , L21812. doi:10.1029/2005GL024584.

    • Search Google Scholar
    • Export Citation
  • Samad, T., and S. A. Harp, 1992: Self-organization with partial data. Network-Comput. Neural Syst., 3 , 205212.

  • Simmonds, I., and K. Keay, 2000: Mean Southern Hemisphere extratropical cyclone behavior in the 40-year NCEP–NCAR reanalysis. J. Climate, 13 , 873885.

    • Search Google Scholar
    • Export Citation
  • Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131 , 29613012.

  • Verdon-Kidd, D., and A. S. Kiem, 2008: On the relationship between large-scale climate modes and regional synoptic patterns that drive Victorian rainfall. Hydrol. Earth Syst. Sci. Discuss., 5 , 27912815.

    • Search Google Scholar
    • Export Citation
  • Wan, H., X. L. Wang, and V. R. Swail, 2007: A quality assurance system for Canadian hourly pressure data. J. Appl. Meteor. Climatol., 46 , 18041817.

    • Search Google Scholar
    • Export Citation
  • Wang, X. L. L., 2008: Accounting for autocorrelation in detecting mean shifts in climate data series using the penalized maximal t or F test. J. Appl. Meteor. Climatol., 47 , 24232444.

    • Search Google Scholar
    • Export Citation
  • Wang, X. L., V. R. Swail, and F. W. Zwiers, 2006: Climatology and changes of extra-tropical cyclone activity: Comparison of ERA-40 with NCEP–NCAR reanalysis for 1958-2001. J. Climate, 19 , 31453166.

    • Search Google Scholar
    • Export Citation
  • Wang, X. L., F. W. Zwiers, V. R. Swail, and Y. Feng, 2009: Trends and variability of storminess in the northeast Atlantic region, 1874-2007. Climate Dyn., 33 , 11791195. doi:10.1007/s00382-008-0504-5.

    • Search Google Scholar
    • Export Citation
  • WASA Group, 1998: Changing waves and storms in the northeast Atlantic? Bull. Amer. Meteor. Soc., 79 , 741760.

  • Wijngaard, J. B., A. M. G. Klein Tank, and G. P. Können, 2003: Homogeneity of 20th century European daily temperature and precipitation series. Int. J. Climatol., 23 , 679692.

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

    Location of stations (⋄) used in this study. Some metropolitan centers (x) are also marked for reference. Macquarie Island (54.5°S, 158.94°E) and Heard Island (53.02°S, 73.39°E) were also digitized but are not marked on the map.

  • Fig. 2.

    Winter (JJA) synoptic patterns for Australia derived from ERA-40 data between 1958 and 2001 (SOMERA). The numbers in brackets indicate the relative frequency of the associated SOMOBS nodes between 1907 and 2006 (see text for details). Decadal trends m in the SOMOBS node frequency are also shown. An asterisk after the number indicates that the trend is statistically significant at the 5% level using a nonparametric Mann–Kendall test.

  • Fig. 3.

    Frequency of each SOM node from Fig. 2 from 1907 to 2006. The thick gray line indicates decadal variations in the data using a 21-term binomial filter. Correlations calculated over the period from 1907 to 2005 are also shown between the frequencies of each node and the JJA SAM index (Jones et al. 2009) and the seasonally averaged JJA Niño-4 index calculated from HadISST1 (Rayner et al. 2003).

  • Fig. 4.

    (a) Bars represent the percentage of synoptic patterns occurring during winter “severe storms” at Cape Otway overlaid with the relative frequency of each node within the SOM; (b) the ratio of when synoptic patterns are driving winter severe storms at Cape Otway to the relative frequency of that pattern within the SOM is represented. The x axis represents the SOM nodes from Fig. 2.

  • Fig. 5.

    Time series of total rainfall for Perth associated with each of the SOM nodes from Fig. 2. Decadal trends m are also given (mm). The line of best fit using an ordinary least squares regression (solid and dashed lines) and that these trends are significant at the 5% level using a Mann–Kendall test is indicated (solid lines).

  • Fig. 6.

    The first (second) box plot in each panel represents winter (JJA) daily rainfall intensity from rain days (≥1 mm) during SOM nodes that are decreasing (increasing) in frequency. Data used from stations in the high-quality dataset (Haylock and Nicholls 2000) closest to (a) Sydney, (b) Melbourne, (c) Adelaide, and (d) Perth using all daily data from 1910 to 2005.

  • Fig. 7.

    Rainfall totals (PRCPTOT), maximum daily rainfall (RX1day), and daily rainfall intensity (SDII) associated with node 13 from Fig. 2 at four Australian cities. Decadal trends m are also given (mm, PRCPTOT and RX1day; and mm day−1, SDII). The line of best fit using an ordinary least squares regression (straight solid and dashed line) and that these trends are significant at the 5% level using a Mann-Kendall test (solid lines) are indicated.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 699 154 22
PDF Downloads 307 46 3