1. Introduction
The study of clouds and their associated precipitation is a cornerstone of atmospheric research, of which applications such as weather and climate prediction provide much needed information about trends and spatiotemporal variability of rainfall. One cloud feature, the Australian northwest cloudband, is notable for its striking appearance and has gained research interest for its frequency of occurrence and association with widespread continental rainfall. The amount of seasonal rain attributed to cloudband events varies greatly by region (Reid et al. 2019; Wright 1997) with highest impact generally found in central Australia and agriculturally marginal areas. Notably, Wright (1997) attributed up to 90% of April–October rainfall in north western and central Australia to northwest cloudbands.
Like other recurrent cloudbands (e.g., the South Pacific and South Atlantic convergence zones) the Australian northwest cloudband extends from the tropics to the midlatitudes. The broader circulation associated with it provides a channel for latent heat and moisture transport. Observational studies of singular cloudband events have sought to answer how atmospheric conditions support cloudband development. The combination of warm ocean surface temperatures and a strong subtropical jet situated over northern Australia have been shown to support late autumn and winter cloudbands (e.g., Gentilli 1974). The amplitude of the midlatitude trough over Western Australia (WA) and preexisting baroclinity over the continent was shown to support a cloudband with extreme rainfall (Downey et al. 1981). Frederiksen and Frederiksen (1996) used a simulated mean winter circulation to find changes associated with cloudband development that agreed well with previous observational studies; highlighted in that study was an equatorward shift of the midlatitude storm track and enhancement of the subtropical jet.
The Australian Bureau of Meteorology (BoM) includes the northwest cloudband among other climate influences such as El Niño–Southern Oscillation (ENSO), Madden–Julian oscillation (MJO), and southern annular mode (SAM) (Bureau of Meteorology 2020). According to the BoM and other sources, rainfall is attributed both to the cloudband itself and to interactions between the cloudband and cold fronts or cutoff lows (e.g., Wright 1997; Dowdy and Catto 2017). The Indian Ocean dipole (IOD) is another climate influencer that has steadily gained research interest due to its potential impact on drought in Australia (Ummenhofer et al. 2009; Verdon-Kidd and Kiem 2009) and conversely heavy rain events (Ashok et al. 2003). Work by Streten (1983) and Nicholls (1989) investigated Indian Ocean sea surface temperature (SST) anomalies and their impact on Australian rainfall, speculating a connection to the appearance of northwest cloudbands. More recently, Reid et al. (2019) suggested the SST gradient across the Indonesian Archipelago and Indian Ocean to be a key driver of winter cloudbands.
Previous studies of Australian northwest cloudband events have left room for improvement, both in the construction of a decades-long climatology and in the investigation of cloudband dynamics. A particular gap in the study of cloudbands remains in the modeling of these events, for which this study provides a novel approach with the use of a long control simulation from a coupled climate model. The aim of this study has four major components to address these gaps in our understanding of cloudbands and their associated rainfall:
Develop an improved 40-yr climatology of northwest cloudband events using a jet-based search algorithm that we have designed to reduce the occurrence of “false alarm” events.
Use this cloudband climatology to strategically investigate sources of moisture for its rainfall, the dynamical structure of the cloudband, and its relationship to the larger-scale flow.
Use this improved climatology of cloudbands to determine what proportion of rainfall across the continent is due to this cloud feature.
Investigate the ability of a coupled climate model to simulate northwest cloudbands, their dynamics, and long period variability.
2. Methods
a. Data
A variety of data sources were used to compute the key atmospheric features needed to detect Australian northwest cloudbands and to interrogate the dynamics associated with their presence. The Japanese 55-year Reanalysis (JRA-55) (Kobayashi et al. 2015) provides atmospheric data from 1958, of which a subset of daily data from 1979 to 2018 at a horizontal grid spacing of 1.25° were used in this study. All anomaly fields computed with JRA-55 were computed using the 1979–2018 period as the daily climatological mean. Unless specified otherwise, the following quantities were computed using JRA-55.
Total wind speed [(u2 + υ2)1/2] at 300 hPa was computed to represent the subtropical jet stream. Geopotential height anomalies at 850 and 500 hPa were computed to represent the atmospheric flow and indicate anomalous high (positive) and low (negative) pressure systems.
Outgoing longwave radiation (OLR) anomalies from 1979 to 2018 were computed by the difference of daily values and their long-term means. Daily OLR data were acquired by satellite and provided with temporal interpolation by NOAA/OAR/ESRL Physical Sciences Laboratory (https://psl.noaa.gov/) available on a 2.5° × 2.5° global grid (Liebmann and Smith 1996). Long-term means for daily values of OLR were computed using the period 1981–2010. Negative OLR anomalies were used to represent cloudiness.
The Optimum Interpolation Sea Surface Temperature (OISST) v2 high-resolution dataset (Reynolds et al. 2007) provides SST on a 0.25° global grid, of which a subset from 1993 to 2017 reinterpolated to the JRA-55 1.25° resolution was computed and used for this study. SST anomalies were computed using the same period 1993–2017, as the climatological mean. Satellite SST sensors involved in the production of this OISST reanalysis were the Advanced Very High Resolution Radiometer (AVHRR) and Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) available from 2002 to 2011.
The Australian Water Availability Project (AWAP) provides high-resolution, quality rainfall data over the Australian continent at a 5-km spatial resolution (Jones et al. 2009; Raupach et al. 2009). AWAP daily rainfall totals, represented by the total precipitation (including rain, snow, hail and dew) accumulated in the 24-h period to 0900 local time, were selected from 1979 to 2017 and reinterpolated to a 0.5° resolution for analysis.
One goal of this study was to understand cloudband dynamics both through investigation of atmospheric reanalysis products and in a coupled climate model where radiative forcing is held constant. Evaluating how well a general circulation model (GCM) can simulate the atmospheric processes involved in extreme events (like the rainfall associated with cloudbands) is referred to as process evaluation and can shed light on why a GCM may or may not simulate the correct frequency or magnitude of extreme events (e.g., Tozer et al. 2020). Model evaluation helps to assess model skill and to see patterns and connections not easily seen with much shorter, observed or reanalysis datasets. For this purpose, 300 years of a long control simulation of the ACCESS-D coupled model (O’Kane et al. 2021a,b) were investigated and compared to the frequency, climatology, and dynamical behavior of cloudbands in the reanalysis. ACCESS-D is built upon the Geophysical Fluid Dynamics Laboratory (GFDL) CM2.1 coupled climate model (Delworth et al. 2006). From the model, with an atmospheric spatial resolution of 2.5° longitude × 2° latitude, the following features were computed in the same manner as with JRA-55: total wind magnitude at 300 hPa, geopotential height anomalies at 500 hPa, wave activity flux at 500 hPa, and integrated water vapor transport. It is worth noting here that the model’s coarse spatial resolution in comparison to the reanalysis likely affects cloudband detection, though the aim in this study is to assess the model in its current resolution.
b. NW cloudband detection algorithm
Previous studies of cloudbands have categorized events in a variety of ways, often with slightly different terminology for those cloudbands with a northwest–southeast orientation across Australia (e.g., Australian northwest cloudbands (acronymized as NWCBs) (Reid et al. 2019), northwest oceanic tropical–extratropical cloudbands (Wright 1997), northwest Australian cloud band (Tapp and Barrell 1984). To note, this study is focusing exclusively on Australian northwest cloudbands and will be referred to as such, or simply as “cloudbands” or “northwest cloudbands.”
Previous studies have investigated various satellite products to search for cloudbands using manual, and more recently, automated detection methods. Infrared and visible imagery were used in pioneering studies (e.g., Tapp and Barrell 1984; Kuhnel 1990; Wright 1997), while more recent efforts have included brightness temperature (e.g., Fröhlich et al. 2013) and OLR and total cloud cover (e.g., Telcik and Pattiaratchi 2014). Australian northwest cloudbands have historically included the following well-established criteria to define an event:
Cloudband must exist over part of continental Australia.
Cloudband must be spatially continuous.
Cloudband must form between 0°–20°S, 80°–130°E.
Cloudband must have a northwest to southeast orientation.
Cloudband must have a length of at least 2000 km.
Jet-based search algorithm to detect northwest cloudband events.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
Example of jet-based search algorithm detection of a northwest cloudband event on 11 Jul 2016. (top) Himawari-8 visible satellite imagery. Shaded values indicate (bottom left) 300-hPa wind magnitude and (bottom right) OLR daily mean anomalies. The stippled boxes in the bottom plots mark the location of the OLR search box determined by characteristics of the subtropical jet (jet core and tilt marked in the bottom-left plot by red and yellow lines, respectively).
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
Extending the cloudband dataset for further analysis is advantageous for several reasons. One reason is evident from Kuhnel’s (1990) 5-yr cloudband climatology over the period 1979–83 which coincided with a strongly positive phase of ENSO. Analysis results from this period were noted to be likely biased toward this mode of variability. Wright (1997) formed an archive of cloudband events over the period 1978–92, including only the cool season months of April–October. While the cool season does encapsulate the peak of cloudband activity, truncating in this manner may create biases in our understanding and interpretation of cloudband dynamics. Recently, work by Reid et al. (2019) has gone into extending the cloudband dataset for a 30-yr period over 1984–2014 using satellite-based detection. Here we further increase to a 40-yr period over 1979–2018 using reanalysis and satellite OLR.
3. Results
a. Cloudband detection algorithm verification
The year 2016 was chosen to perform a verification of the performance of the jet-based cloudband search algorithm, due to the large number of events that occurred this year (see Fig. 4). A contingency table is one method to perform verification of a dichotomous (yes or no) forecast that can be applied to see what types of errors are being made. Here, our “forecast” value is derived from the search algorithm result and our “observed” value is derived from a manual detection of cloudbands using 6-hourly Himawari-8 visible satellite imagery [Himawari-8/-9 gridded data are distributed by the Center for Environmental Remote Sensing (CEReS), Chiba University, Japan] using the historical cloudband criteria detailed in the methods section. A cloudband day was recorded if any of the four 6-hourly time steps in that day met all cloudband criteria. A successful algorithm will have a high number of hits (both algorithm and manual search say “yes”) and correct negatives (both algorithm and manual search say “no”) and a low number of misses (algorithm says “no,” manual search says “yes”) and false alarms (algorithm says “yes,” manual search says “no”).
The contingency table for the 2016 verification of daily northwest cloudbands detected by the jet-based search algorithm is presented in Fig. 3. The algorithm scored an accuracy [(hits + correct negatives)/total)] of 0.78 and a probability of detection [hits/(hits + misses)] of 0.52. A false alarm ratio [false alarms/(hits + false alarms)] measured 0.36 with a probability of false detection [false alarms/(correct negatives + false alarms)] of 0.12. Missed events generally occurred when the jet criterion were too weak to satisfy the algorithm, while false alarms occurred mainly as a result of broken but extensive cloud cover that failed to satisfy the “cloudband must be continuous” criteria. While not pursued directly in this study, these false alarm events may have been cloudband events in the decay stage. Skill was also evaluated by season. Winter (JJA) had the highest proportion of false alarms (0.42) with a lower proportion of misses (0.2) compared to all other seasons, resulting in the highest probability of false detection (0.64). Conversely, summer (DJF) had the lowest probability of detection (0.36). The highest probability of detection (0.66) occurred in spring (SON).
(left) Contingency table for 2016 verification of daily northwest cloudbands detected by jet-based search algorithm. Here, “manual” refers to visual inspection of Himawari-8 visible satellite imagery. (right) Distributions of AWAP rainfall between 20° and 40°S for each event verification type are represented by boxes extending from the lower to upper quartile values with a line at the median. Whiskers show the range of the data, beyond which outliers are excluded.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
To further reassure that the inclusion of false alarm cloudband events was not detrimental to composite analysis, rainfall distributions were compared between all contingency types (Fig. 3). The distribution of false-alarm-type rainfall most closely matched the distribution of hit (true event)-type rainfall, indicating that the large-scale atmospheric drivers of both event types were similar. Conversely, missed events had a rainfall distribution similar to that for correct negative events. Taken together, the four verification type rainfall distributions indicated that the jet-based search algorithm was successful in capturing the rainfall signature of true cloudband events and nonevents.
From the verification of cloudbands detected using this new algorithm, we are confident that the dynamically based searching method is able to produce a high-quality cloudband dataset while simultaneously keeping the probability of false detection low in most seasons.
b. 40-yr cloudband climatology
Progressing from the verification of the algorithm-detected cloudbands in 2016, the 40-yr period from 1979 to 2018 was next analyzed. Details of the yearly and monthly variability of algorithm-detected northwest cloudbands during this period are visualized in Fig. 4. Each black bar within the top plot of Fig. 4 represents one northwest cloudband day, so that the summation of columns yields 1979–2018 daily totals and summation of rows yields each year’s event total. A total of 1448 cloudband days were detected during this period, with a maximum of 86 cloudband days in 2016, and a minimum of 11 cloudband days in 1982 and 2006. The average number of cloudband days per year was 36. Considering consecutive cloudband days to be one event’s lifespan, the average duration of a cloudband’s lifetime was 1–2 days with the longest lifespan being 8 days.
A 40-yr (1979–2018) climatology of algorithm-detected Australian northwest cloudband events. (bottom left) Daily and (bottom right) yearly cloudband event sums are provided. (top) Each individual cloudband event is indicated by a black bar that can be summed vertically to the January–December daily time series and summed horizontally to the 1979–2018 yearly time series.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
A pronounced seasonal cycle was evident that featured an increase in cloudband activity during April–June, peaking in May. Cloudband activity has a tendency to then decrease in the late winter before rising again in October. A lull in cloudband activity between January and March is consistent with some of the earlier investigations of cloudband seasonality (Tapp and Barrell 1984; Kuhnel 1990) and with the Australian BoM’s current guidance on cloudband influence (Bureau of Meteorology 2013). A discussion on how this result compares to studies that found a frequency peak of cloudbands in warmer months (e.g., Fröhlich et al. 2013; Reid et al. 2019) is provided in the summary and conclusions section of this paper. No significant trend could be established of annual cloudband activity over the period 1979–2018. However, there is clear evidence of a decrease in cloudband activity over the period 2001–09, which falls within the period known as the Millennium drought. This is evidenced in Fig. 5 by an average yearly cloudband total between 2001 and 2009 that falls below the lower quartile of the entire 1979–2018 dataset. Four of the years in the 2001–09 period have the four lowest numbers of cloudbands in the record. Considered to be one of Australia’s most severe droughts, the Millennium drought was a series of concurrent severe droughts between the mid-1990s and late 2000s attributed to a range of different climates modes (Risbey et al. 2013; Verdon-Kidd and Kiem 2009) that was broken by a wet year in 2010 (Cai et al. 2014). The Millennium drought break is reflected here by a sharp increase in cloudband activity in 2010. Of note, the decrease in cloudband activity during the Millennium drought is most evident during the primary peak months of April–June, and not within the secondary peak of austral spring/early summer.
Distribution of 1979–2018 yearly cloudband sums (boxplot extending from the lower to upper quartile values with a line at the median) with individual (pink markers) and average (red marker with black outline) 2001–09 yearly cloudband sums compared to the individual (teal markers) and average (cyan marker with black outline) 1979–2000 + 2010–18 yearly cloudband sums.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
c. Atmospheric diagnostic composites
Following development and verification of the search algorithm and 40-yr dataset of Australian northwest cloudbands, we now seek to understand the dynamical structure of the cloudband and its relationship to the large-scale atmospheric flow. Composite analysis works well to reveal common large-scale features at the expense of smoothing smaller-scale features that may contribute to cloudband development. Nevertheless, with a sample size of 1448 cloudband days and Monte Carlo significance testing at the 95th percentile (not shown in this paper), we can be confident that the atmospheric features revealed by composite analysis and discussed herein are unique to the northwest cloudband case.
As early as Streten (1973), it has been conjectured that the location of midlatitude cloudbands is related to the longwave hemispheric pattern. More recently, Reid et al. (2019) used composite analysis spanning 30 years of winter (JJA) northwest cloudband days to illustrate an anomalous low pressure trough at 750 hPa and centered over continental Australia, which tilted westward with height up to 250 hPa. In this study, composite 500-hPa geopotential height anomalies revealed an anomalous low pressure trough tilted in a northwest–southeast orientation positioned over southeast Australia (Fig. 6). This tilted trough was embedded in a Rossby wave train beginning around 60°E and ending around 120°W. This trough reinforced the upper-level subtropical jet stream (northern branch of gray shading in Fig. 6), which also had a composite northwest–southeast orientation. Along the polar jet stream (southern branch of gray shading in Fig. 6), wave activity flux vectors at the 500-hPa level revealed a channel of wave packet propagation with an excursion of wave activity directed into the anomalous midlatitude trough. Taken together, the diagnostics in Fig. 6 illustrate that the dynamical structure of the Australian northwest cloudband is related to and reinforced by the large-scale upper-level flow through a Rossby wave train.
Composites of all 1979–2018 algorithm-detected Australian northwest cloudband events for 500-hPa geopotential height anomalies [black contours, solid (dashed) lines indicate positive (negative) anomalies], 500-hPa wave activity flux (purple vectors, masked to the 85th percentile), and 300-hPa wind magnitude (gray shading of >25, 30, and 35 m s−1).
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
Looking lower, Fig. 7 presents the anomaly composites of sea surface temperature (SST), thermal wind, and 850-hPa geopotential height. The 850-hPa low pressure trough embedded within a wave train is collocated with the 500-hPa tilted trough described above. Composite SST anomalies reveal several distinct ocean surface signatures. Cooler than normal tropical Pacific SSTs resembled a central Pacific La Niña–like pattern, while warm surface temperatures around the Maritime Continent are juxtaposed with cooler SST off the coast of WA. Little signal was revealed in the western portion of the Indian Ocean. The alternating cool–warm SST anomalies centered around the 45°S latitude band are consistent as the surface signature of the atmospheric wave train.
Composites of all 1979–2018 algorithm-detected Australian northwest cloudband events for 850-hPa geopotential height anomalies [gray contours, solid (dashed) lines indicate positive (negative) anomalies], thermal wind anomalies [black contours, solid (dashed) lines indicate positive (negative) anomalies], and SST anomalies (blue to red shading). Colored boxes mark regions for calculating the Niño-3.4 index (solid red), DMI index (difference of solid blue and dashed blue boxes), and DMI EAST index (dashed cyan box). Black dots mark locations used to calculate the Indian Ocean SST gradient as discussed in the text.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
The composite shows a “train” of thermal wind anomalies set up over the Australian continent with a strong positive anomaly positioned along the axis of the gradient of cool to warm SST anomalies in the eastern Indian ocean. Regarding whether the dipole-like SST anomaly set up the thermal wind anomaly response in the atmosphere (or vice versa), we note that the SST anomaly signature was present out to 7 days prior to a coherent thermal wind anomaly and development of the cloudband (not shown). Simmonds (1990) also similarly found an ocean to atmosphere response through the SST gradient to the northwest of the Australian continent that set up the atmospheric circulation conducive to winter precipitation. Further, the thermal wind anomaly is dynamically consistent with the sign and location of the warm (northern)–cold (southern) SST anomaly field. This temperature gradient would imply a NW–SE-directed enhancement of the thermal wind and hence subtropical jet in the location where it influences the cloudband. It has been additionally demonstrated that over the region where our anomalies manifest, ocean-to-atmosphere predictability is stronger than atmosphere-to-ocean predictability (Bach et al. 2019). Taken together, the compositing result here suggests that the SST gradient pattern is enhancing the thermal wind and hence the tilted subtropical jet, favoring development of the tilted trough over the continent.
Distinct SST anomaly signatures in the tropical Pacific and Indian Oceans raise the question of the influence of ENSO and IOD on Australian northwest cloudband activity. Convective activity over the northwest cloudband region has mainly been attributed to the atmospheric dynamical mechanisms that excite ENSO and IOD events (rather than the ocean-atmosphere response to warm SSTs) (Webster et al. 1998); however, both ENSO and IOD have been attributed as drivers themselves for cloudband variability (Telcik and Pattiaratchi 2014; Nicholls 1989; Reid et al. 2019).
Presented here for the first time we see a long-term climatology of observed cloudband events associated with strong La Niña presence and anomalously warm ocean SSTs over the Maritime Continent. This dual-signature can be investigated by the amount of cloudband days that occur while different climate indices are positive, neutral, and negative. Figure 8 supplies the number of cloudband days in the analysis period that occur during different phases of four SST anomaly indices (regions to calculate these indices are provided as a reference in Fig. 7). The Niño-3.4 index (computed as the average SST anomaly across 5°N–5°S, 170°–120°W) was selected to represent ENSO and was used as a base comparison to three SST anomaly indices across the Indian Ocean. The Dipole Mode Index (DMI) was selected to represent the intensity of the IOD and computed by the anomalous SST gradient between the western equatorial Indian Ocean (10°S–10°N, 50°–70°E) and the southeastern equatorial Indian Ocean (10°S–0°, 90°–110°E). The DMI EAST index was computed using only the south eastern equatorial Indian Ocean (10°S–0°, 90°–110°E) region. A third index, here referred to as the gradient index, was computed as the difference in SST anomaly in two locations (marked in Fig. 7), chosen to span the gradient of warm to cool SST anomalies in the cloudband composite. Values in each box are the number of cloudband events scaled by the number of months that contributed to that box’s total. Coloring helps to highlight that when the DMI is negative, cloudband events occur over all phases of Niño-3.4 and favor neutral and positive phases. That a negative DMI–positive Niño-3.4 combination appears to dominate is likely an outlier effect of a small sample size for this combination. A higher number of cloudband events occur with simultaneous negative Niño-3.4 and positive DMI EAST–gradient index values. This matched our results seen in Fig. 7. The reduced concurrency of the full DMI positive phase with Niño-3.4 is unsurprising due to the lack of any anomalous SST over the western equatorial Indian Ocean.
Number of cloudband days over the period 1979–2018 that occur during phases of Niño-3.4 compared to the (left) DMI, (center) DMI EAST, and (right) Indian Ocean gradient indices. Values are the number of cloudband events scaled by the number of months that contributed to that box’s total, emphasized by coloring.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
d. Rainfall and moisture
The association between Australian northwest cloudbands and rainfall has long been recognized; in this study we look at this association from a few new angles. Rainfall deciles provide a meaningful illustration of rainfall contribution by damping the signal of areas that receive high yearly mean rainfall regardless of the driving synoptic mechanism. “Decile” here refers to percentiles of the dataset that are divided by 10 successive tenths from 0 to 100. The decile of composite rainfall for cloudband days was determined from the distribution of rainfall data across Australia over the period 1979–2017 using AWAP data. From Fig. 9 it becomes clear that the rainfall delivered by northwest cloudbands contributes to the wettest conditions over the 40-yr analysis period in central interior Australia extending northwestward to the WA coast. Strikingly, almost all locations across Australia fall within the upper decile ranking of composite cloudband day rainfall. Only a few locations fall within the average decile ranking (far northern portions of Australia, western Tasmania) or where decile information is unreliable due to data sparseness. Another way to assess the importance of cloudband rainfall is to compute the fraction of total rainfall over the analysis period that falls on cloudband days. Similar to the composite rainfall deciles, Fig. 10 shows that rainfall attributed to northwest cloudbands contributed significantly to total rainfall in central Australia and along the WA coast. Over 1979–2017, areas of interior Australia received as much as a quarter of their total rainfall from cloudband events.
AWAP rainfall deciles for all 1979–2017 algorithm-detected Australian northwest cloudband event days. Gray shading indicates areas where data are too sparse to compute a reliable decile value.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
Fraction of total 1979–2017 rainfall attributed to Australian northwest cloudbands. White shading indicates area masked out due to data unavailability.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
The source of moisture to support cloudband rainfall has not been rigorously investigated. Some assert that this moisture transport is directed along the cloudband and upper-level atmospheric flow from over warm Indian Ocean waters (Bureau of Meteorology 2013). It is judicious to test this assumption. The NOAA HYSPLIT trajectory model (Stein et al. 2015; Draxler and Hess 1997) was used to determine the origin of air parcels located with respect to the subtropical jet center during cloudband events in 2016 (Fig. 11). Parcels were seeded at 700 hPa, which is the approximate level where available water droplets undergo growth by condensation and collision–coalescence to fall as rain. Each parcel was traced back ten days from the start of a cloudband event. Results from this analysis indicate that when a cloudband event is detected with a jet center west of 130°E, air parcels equatorward of the jet have arrived predominately from the Timor Sea region. In comparison, air parcels equatorward of the jet centered east of 130°E have arrived from a spread of Maritime Continent waters including the Timor, Arafura, and Coral Seas. Regardless of jet center location, air parcels poleward of the jet originate over colder, higher-latitude Southern Ocean water. This result indicates that the dynamical moisture transport associated with cloudband rainfall is not singularly tied to the Indian Ocean. Recent work using an alternative back-tracking method agrees with our result. Holgate et al. (2020) find that while northwest cloudbands have been linked to precipitation in southeast Australia (Bureau of Meteorology 2013), their back-tracking results show no evidence that precipitation in southeast Australia relies on moisture sourced from the northwest of Australia.
Location of parcel back trajectories using the NOAA HYSPLIT trajectory model for 2016 cloudband events. Colors of dots indicate where parcels were seeded in each case with respect to the location of the atmospheric subtropical jet center (legend on right). Color transparency indicates time from event origin [least (most) transparent equals day 0 (day −10)]. All parcel backtracks were seeded at 700 hPa.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
To further understand this dynamical moisture transport we explore the influence that ARs may play in cloudband rainfall development. As discussed in the methods section of this paper, integrated water vapor transport (IVT) is used as a standard metric of ARs. Figure 12 presents the composite of IVT for all 1979–2018 algorithm-detected northwest cloudband days. IVT on each cloudband day was first aligned to the jet center and then composited. This jet-centering composite technique revealed an AR directed from the equator poleward and northeast of the jet center. This AR was not in direct alignment with the composite subtropical jet (see northern branch of gray shading in Fig. 6, dashed contour in Fig. 12). A more nearly N–S orientation directs the transport of moisture from the tropics into the vicinity of the cloudband. Combined, the HYSPLIT backtracks and IVT composite indicate that water vapor originating north of the cloudband, and not northwestward along the cloudband, is the dominate supplier of moisture to support rainfall during cloudband events.
Composite of IVT over all 1979–2018 algorithm-detected Australian northwest cloudband events, aligned to the jet center of each case. The black square marks the location to where all events were centered. The dashed line indicates the 35 m s−1 contour of the composite 300-hPa wind magnitude, aligned to the black square (jet center) in the same manner as IVT.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
e. 300-yr ACCESS-D cloudband climatology
The remainder of this study assesses the climatology and behavior of Australian northwest cloudbands within ACCESS-D, a 300-yr coupled control simulation described in the methods section. After a 1500-yr model spinup to stabilize internal drift, years 1500–1799 were analyzed (note that these years are model years and not related to any historical record). Figure 13 presents the ACCESS-D cloudband climatology in the same manner as Fig. 4. Black bars represent one northwest cloudband day, so that the summation of columns yields the 300-yr daily totals and summation of rows yields each year’s event total. A total of 2659 cloudband days were detected during this period, with a maximum of 25 cloudband days occurring within one year. A minimum of zero cloudband days was recorded in 2 of the 300 years. The average number of cloudband days per year was nine; this was significantly fewer than the 36 yr−1 average in our observational climatology. Considering consecutive cloudband days to be one event’s lifespan, the average duration of a cloudband’s lifetime was 1 day, with a 6-day maximum duration. This was similar but slightly shorter than observed cloudband duration. The seasonal cycle featured a primary peak in cloudband activity in September, preceded by a minimum in February and generally increasing trend over late austral autumn/winter. No significant annual or multidecadal trend was observed in cloudband activity using the 300-yr control simulation, which is unsurprising given the long spinup and steady forcing of the control run.
300-yr climatology of algorithm-detected Australian northwest cloudband events in the ACCESS-D control run. (bottom left) Daily and (bottom right) yearly cloudband event sums. (top) Each individual cloudband event is indicated by a black bar that can be summed vertically to the January–December daily time series and summed horizontally to the model year 1500–1799 yearly time series.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
A comparison of cloudbands detected using the 40-yr reanalysis dataset to the 300-yr control run dataset is provided in Fig. 14. Prominently evident are the difference in both cloudband seasonal cycle and monthly event distribution. While the February minimum is consistent, the primary peak of ACCESS-D cloudbands is around September compared to the reanalysis cloudband peak in May. With more than 7 times the amount of years included in the control run versus the reanalysis, the distribution of monthly sums is lower for each month of the year for the former compared to the latter.
Comparison of Australian northwest cloudbands in the JRA-55 and ACCESS-D control run. (top) Monthly sums of cloudbands for JRA-55 (blue) and ACCESS-D (orange) for 40 and 300 years, respectively. (bottom) The distribution of monthly cloudband sums.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
We next investigate if the upper-level dynamics and moisture transport from a long record of model cloudbands reflect our observational analysis. From Fig. 15 we see a similar tilted trough embedded in an anomalous Rossby wave train at 500 hPa from a composite of all 2659 cloudband days detected in the control run. The anomalous low center was shifted farther south and organization of the Rossby wave train extended beyond the longitude bounds of the reanalysis composite wave train. No determinable southward shift of the Hadley cell occurs in the model to explain this feature (Fig. S2). Both composite 500-hPa wave activity flux and subtropical jet stream compared well between the control run and reanalysis.
Composites of 300 years of ACCESS-D control run algorithm-detected Australian northwest cloudband events for 500-hPa geopotential height anomalies [black contours, solid (dashed) lines indicate positive (negative) anomalies], 500-hPa wave activity flux (purple vectors, masked to the 85th percentile), and 300-hPa wind magnitude (gray shading of >25, 30, and 35 m s−1).
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
Compositing SST anomalies of model cloudbands proved trickier as ocean variables at the time of analysis were only available in monthly averages. To provide a closer “apples-to-apples” comparison of SST anomalies to daily composite thermal wind anomalies, only months in the model record with eight or more cloudband days (the upper 99th percentile of monthly cloudband sums) were used to composite SST anomalies. This method provided 27 cloudband events over 3 months as our composite sample. The resulting composite is shown in Fig. 16. Within a noisier composite we a similar warm to cool SST anomaly gradient present off the coast of WA in the Indian Ocean, as compared to the reanalysis. This again coincides with a train of thermal wind anomalies set up over the Australian continent with a strong positive anomaly positioned along the axis of the SST anomaly gradient. It is noted that the ACCESS-D model was shown to have reasonable and well-distributed ENSO behavior (as shown in Fig. S3).
Composites of 300 years of ACCESS-D control run algorithm-detected Australian northwest cloudband events for thermal wind anomalies [black contours, solid (dashed) lines indicate positive (negative) anomalies] and SST anomalies (blue to red shading). Colored boxes mark regions for calculating the Niño-3.4 index (solid red), DMI index (difference of solid blue and dashed blue boxes), and DMI EAST index (dashed cyan box). Black dots mark locations used to calculate the Indian Ocean SST gradient as discussed in the text.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
The rainfall attributed to the control run’s northwest cloudbands reflects the composite synoptic features, which are similar to the reanalysis but shifted poleward (Fig. 17, note the change in scale). The poleward displacement of the anomalous upper-air trough in the model shifts the highest fraction of total rainfall due to cloudbands poleward of the observed counterpoint. The model’s composite IVT pattern (Fig. 18) reproduced the AR directed from the equator poleward seen in the reanalysis composite (though the model AR is less connected to its northern counterparts). With fewer cloudbands detected per model year, the model still reproduced the observed dynamical structure of cloudband reinforcement via the large-scale upper-level flow through an anomalous Rossby wave train and rainfall support via AR moisture transport. Our composite analysis results are an encouraging indicator of the control run’s ability to simulate Australian northwest cloudband dynamics.
Fraction of total rainfall attributed to 300 years of ACCESS-D control run algorithm-detected Australian northwest cloudbands.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
Composite of IVT over 300 years of ACCESS-D control run algorithm-detected Australian northwest cloudband events, aligned to the jet center of each case. The black square marks the location to where all events were centered.
Citation: Monthly Weather Review 149, 4; 10.1175/MWR-D-20-0308.1
4. Summary and conclusions
A dynamical search algorithm was developed to create a 40-yr climatology of northwest cloudband events. Verification of one year of this cloudband dataset provided confidence that “false alarm” events were either limited or not detrimental to composite analysis results based on the similar distribution of false alarm-type rainfall and verified cloudband rainfall. Manual inspection of cloud satellite imagery further supported the inclusion of false alarm-type events into the set of events by recognition that many of these events involved a cloudband in a decaying state that did not satisfy the “spatially continuous” criterion. A pronounced seasonal cycle in events was consistent with previous studies (Tapp and Barrell 1984; Kuhnel 1990), which found cloudband development to peak in early autumn. The range of a cloudband’s lifetime from 1 to 8 days was also consistent with previous studies and with operational guidance about cloudband influence (Bureau of Meteorology 2013). Assessment of this 40-yr cloudband dataset revealed a clear minimum of cloudband activity that coincided with one of Australia’s most well-known dry periods, the Millennium Drought.
Composite analysis of upper- and lower-level diagnostics indicated that 1979–2018 algorithm-detected cloudbands were supported by a few key atmospheric processes. Starting from the ocean surface, a SST anomaly gradient of warm (northern) to cool (southern) waters in the eastern Indian Ocean off the coast of WA was in place a week in advance of a responding thermal wind anomaly. This NW–SE-directed enhancement of the thermal wind corresponded to a tilted subtropical jet, which in turn favored the development of the tilted 500-hPa trough over the continent. This tilted low pressure trough was embedded and supported by a Rossby wave train and strong flux of wave activity at 500 hPa. The significance of other SST signals in the cloudband composite was investigated by comparing the co-occurrence of Niño-3.4 phases to other SST anomaly indices in the Indian Ocean. Cloudband activity was concentrated when Niño-3.4 was in its negative phase (La Niña) and either the DMI EAST box or gradient index (measuring the SST gradient off WA) was strongly positive. Our composite analysis highlights the clear role of anomalously warm Maritime Continent waters to cloudband development.
Australian northwest cloudbands were shown to be associated with extreme rainfall; the rainfall for these events represented the wettest decile conditions of a 40-yr period and impacted regions where yearly rainfall is typically low. Results from the backtracking of parcels located with respect to the subtropical jet for cloudbands in 2016 revealed moisture sources from the Timor, Arafura, and Coral Sea regions. A jet-centered composite of IVT for all 1979–2018 algorithm-detected cloudbands produced a coherent N–S-oriented AR, which would direct the transport of moisture from the tropics into the vicinity of the cloudband. These results indicate that the dynamical moisture transport associated with cloudband rainfall is not singularly tied to the Indian Ocean, as has been previously asserted.
Not all recent cloudband studies share similar seasonal cycle and rainfall patterns with the results presented here. For example, both Fröhlich et al. (2013) and Reid et al. (2019) found the frequency peak of cloudbands to be in the warmer months. It is likely the differences are tied into the methodology of the respective studies and the use of reanalysis versus satellite products to detect the cloud feature. In particular, we emphasize here the importance of the subtropical jet presence to the definition of the cloudband, which may lead to a differing set of northwest cloudbands to previous studies without jet criteria. The identification of frequent cloudbands in the winter months is therefore not biased to the seasonal meridional shift of the jet but rather a natural consequence of this study’s criteria for a cloudband to exist. Reid et al. (2019) produced results of precipitation anomalies associated with cloudbands that reflected the NW–SE orientation of the cloud feature itself, while the rainfall deciles and fraction of rainfall attributed to northwest cloudbands presented here have a more zonal signature. Our result is more comparable to Wright (1997), who showed that the percentage contribution of different classes of cloudbands and their interactions with other rain-producing systems to total April–October rainfall may influence rainfall in a large band extending across interior Australia. We note that the NW–SE orientation of these cloud features does not necessarily translate to a NW–SE rainfall signature, as the rain is associated with the tilted trough, which tracks more nearly west to east across the continent, yielding a more zonal signature.
The long period variability of cloudbands over 300 years of a coupled climate model produced fewer yearly events and missed the peak in cold season events as compared to observed cloudbands evaluated in the reanalysis product. This result is consistent with previous model process evaluation of extreme wet events in Australia, which found atmospheric patterns were well simulated whereas the statistics of extremes were not (Tozer et al. 2020). In principle, the low-frequency behavior of cloudband activity and rainfall should be consistent. Instead, we find that the fraction of total rainfall attributed to cloudbands in the model is greatly reduced (up to 5 times) compared to that of the reanalysis cloudband rainfall fraction. The question of why a long, coupled control simulation produces far fewer cloudband events remains a topic for future research inquiry and could be guided by an investigation of the importance of spatial resolution to cloudband detection.
Analysis of northwest cloudbands and their associated atmospheric processes in a long control run replicated similar composite structures as in the reanalysis cloudband compositing. In particular, a clear Rossby wave train with embedded tilted low pressure trough is situated in conjunction with a tilted subtropical jet. A coherent AR similar to the one found with the reanalysis compositing is also present, which adds confidence to the discussion of moisture sourcing for cloudband associated rainfall. A train of thermal wind anomalies was present over the Australian continent with a strong positive anomaly positioned along the axis of a SST anomaly gradient in the eastern Indian Ocean. This Indian Ocean gradient was a common composite feature of observed and modeled cloudbands and seems to dominate the response of the atmosphere in forming and supporting cloudband development. Better understanding of the causal processes in cloudband events could help direct future research questions surrounding variability of Australian precipitation.
Acknowledgments
This research was supported by the Decadal Climate Forecasting Project at CSIRO. CCC and BMS were also supported by the Centre for Southern Hemisphere Ocean Research and the Earth Systems and Climate Change Hub of the Australian Government’s National Environmental Science Program.
Data availability statement
The reanalysis data used for this work are from the Japanese 55-year Reanalysis (JRA-55) project carried out by the Japan Meteorological Agency (JMA). These data are available at https://jra.kishou.go.jp. Outgoing longwave radiation (OLR) data are from the NOAA/OAR/ESRL Physical Sciences Laboratory. These data are available at https://psl.noaa.gov/. Sea surface temperature (SST) data are from the Optimum Interpolation Sea Surface Temperature (OISST) v2 high-resolution dataset provided by the NOAA/OAR/ESRL Physical Sciences Laboratory. These data are available at https://psl.noaa.gov/. High-quality rainfall data are from the Australian Water Availability Project (AWAP) carried out by the Australian Bureau of Meteorology. Open access to the data is not currently available; contact climatedata@bom.gov.au for further information. Model data are from a long control simulation of the Climate re-Analysis Forecast Ensemble (CAFE) system, also known as ACCESS-D, carried out by the Decadal Climate Forecasting Project at CSIRO. Open access to the data is not currently available.
REFERENCES
Ashok, K., Z. Guan, and T. Yamagata, 2003: Influence of the Indian Ocean Dipole on the Australian winter rainfall. Geophys. Res. Lett., 30, 1821, https://doi.org/10.1029/2003GL017926.
Bach, E., S. Motesharrei, E. Kalnay, and A. Ruiz-Barradas, 2019: Local atmosphere–ocean predictability: Dynamical origins, lead times, and seasonality. J. Climate, 32, 7507–7519, https://doi.org/10.1175/JCLI-D-18-0817.1.
Benedict, J. J., A. C. Clement, and B. Medeiros, 2019: Atmospheric blocking and other large-scale precursor patterns of landfalling atmospheric rivers in the North Pacific: A CESM2 study. J. Geophys. Res. Atmos., 124, 11 330–11 353, https://doi.org/10.1029/2019JD030790.
Branstator, G., 2002: Circumglobal teleconnections, the jet stream waveguide, and the North Atlantic Oscillation. J. Climate, 15, 1893–1910, https://doi.org/10.1175/1520-0442(2002)015<1893:CTTJSW>2.0.CO;2.
Bureau of Meteorology, 2013: Northwest cloudbands. Accessed 15 March 2019, https://www.bom.gov.au/climate/about/australian-climate-influences.shtml?bookmark=nwcloudband.
Bureau of Meteorology, 2020: Australian climate influences. Accessed 20 March 2019, https://www.bom.gov.au/climate/about/australian-climate-influences.shtml.
Cai, W., A. Purich, T. Cowen, P. van Rensch, and E. Weller, 2014: Did climate change-induced rainfall trends contribute to the Australian Millennium drought? J. Climate, 27, 3145–3168, https://doi.org/10.1175/JCLI-D-13-00322.1.
DeFlorio, M. J., and Coauthors, 2019: Experimental subseasonal-to-seasonal (S2S) forecasting of atmospheric rivers over the western United States. J. Geophys. Res. Atmos., 124, 11 242–11 265, https://doi.org/10.1029/2019JD031200.
Delworth, T. L., and Coauthors, 2006: GFDL’s CM2 global coupled climate models. Part I: Formulation and simulation characteristics. J. Climate, 19, 643–674, https://doi.org/10.1175/JCLI3629.1.
Dowdy, A. J., and J. L. Catto, 2017: Extreme weather caused by concurrent cyclone, front and thunderstorm occurrences. Sci. Rep., 7, 40359, https://doi.org/10.1038/srep40359.
Downey, W. K., T. Tsuchiya, and A. J. Schreiner, 1981: Some aspects of a northwestern Australian cloudband. Aust. Meteor. Mag., 29, 99–113.
Draxler, R. R., and G. D. Hess, 1997: Description of the HYSPLIT_4 modeling system. Tech. Rep. NOAA Tech. Memo. ERL ARL-224, NOAA/Air Resources Laboratory, 31 pp.
Feldstein, S., and U. Dayan, 2008: Circumglobal teleconnections and wave packets associated with Israeli winter precipitation. Quart. J. Roy. Meteor. Soc., 134, 455–467, https://doi.org/10.1002/qj.225.
Frederiksen, C. S., and J. S. Frederiksen, 1996: A theoretical model of Australian northwest cloudband disturbances and Southern Hemisphere storm tracks: The role of SST anomalies. J. Atmos. Sci., 53, 1410–1432, https://doi.org/10.1175/1520-0469(1996)053<1410:ATMOAN>2.0.CO;2.
Fröhlich, L., P. Knippertz, A. H. Fink, and E. Hohberger, 2013: An objective climatology of tropical plumes. J. Climate, 26, 5044–5060, https://doi.org/10.1175/JCLI-D-12-00351.1.
Gentilli, J., 1974: Epitropical climatology of the subtropical jetstream. Bonner Meteor. Abh., 17, 137–165.
Holgate, C. M., J. P. Evans, A. I. J. M. van Dijk, A. J. Pitman, and G. D. Virgilio, 2020: Australian precipitation recycling and evaporative source regions. J. Climate, 33, 8721–8735, https://doi.org/10.1175/JCLI-D-19-0926.1.
Jones, D., W. Wang, and R. Fawcett, 2009: High-quality spatial climate data-sets for Australia. Aust. Meteor. Ocean J., 58, 233–248, https://doi.org/10.22499/2.5804.003.
Kobayashi, S., and Coauthors, 2015: The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5–48, https://doi.org/10.2151/jmsj.2015-001.
Kuhnel, I., 1990: Tropical-extratropical cloudbands in the Australian region. Int. J. Climatol., 10, 341–364, https://doi.org/10.1002/joc.3370100403.
Liebmann, B., and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77, 1275–1277.
Mo, R., and H. Lin, 2019: Tropical-mid-latitude interactions: Case study of an inland- penetrating atmospheric river during a major winter storm over North America. Atmos.–Ocean, 57, 208–232, https://doi.org/10.1080/07055900.2019.1617673.
Nicholls, N., 1989: Sea surface temperatures and Australian winter rainfall. J. Climate, 2, 965–973, https://doi.org/10.1175/1520-0442(1989)002<0965:SSTAAW>2.0.CO;2.
O’Kane, T. J., and Coauthors, 2021a: CAFE60v1: A 60-year large ensemble climate reanalysis. Part I: System design, model configuration, and data assimilation. J. Climate, https://doi.org/10.1175/JCLI-D-20-0974.1, in press.
O’Kane, T. J., and Coauthors, 2021b: CAFE60v1: A 60-year large ensemble climate reanalysis. Part II: Evaluation. J. Climate, https://doi.org/10.1175/JCLI-D-20-0518.1, in press.
Ralph, F. M., and Coauthors, 2017: Atmospheric rivers emerge as a global science and applications focus. Bull. Amer. Meteor. Soc., 98, 1969–1973, https://doi.org/10.1175/BAMS-D-16-0262.1.
Raupach, M., P. Briggs, V. Haverd, E. King, M. Paget, and C. Trudinger, 2009: Australian Water Availability Project (AWAP): CSIRO Marine and Atmospheric Research component: Final Report for phase 3. CAWCR Tech. Rep. 13, 67 pp.
Reid, K. J., I. Simmonds, C. L. Vincent, and A. D. King, 2019: The Australian northwest cloudband: Climatology, mechanisms, and association with precipitation. J. Climate, 32, 6665–6684, https://doi.org/10.1175/JCLI-D-19-0031.1.
Reynolds, R. W., T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey, and M. G. Schlax, 2007: Daily high-resolution-blended analyses for sea surface temperature. J. Climate, 20, 5473–5496, https://doi.org/10.1175/2007JCLI1824.1.
Risbey, J., P. McIntosh, and M. Pook, 2013: Synoptic components of rainfall variability and trends in southeast Australia. Int. J. Climatol., 33, 2459–2472, https://doi.org/10.1002/joc.3597.
Simmonds, I., 1990: A modelling study of winter circulation and precipitation anomalies associated with Australian region ocean temperatures. Aust. Meteor. Mag., 38, 151–162.
Stein, A. F., R. R. Daxler, G. D. Rolph, B. J. B. Stunder, M. D. Cohen, and F. Ngan, 2015: NOAA’s HYSPLIT atmospheric transport and dispersion modeling system. Bull. Amer. Meteor. Soc., 96, 2059–2077, https://doi.org/10.1175/BAMS-D-14-00110.1.
Streten, N. A., 1973: Some characteristics of satellite-observed bands of persistent cloudiness over the Southern Hemisphere. Mon. Wea. Rev., 101, 486–495, https://doi.org/10.1175/1520-0493(1973)101<0486:SCOSBO>2.3.CO;2.
Streten, N. A., 1983: Extreme distributions of Australian annual rainfall in relation to sea surface temperature. J. Climatol., 3, 143–153, https://doi.org/10.1002/joc.3370030204.
Takaya, K., and H. Nakamura, 1997: A formulation of a wave-activity flux for stationary Rossby waves on a zonally varying basic flow. Geophys. Res. Lett., 24, 2985–2988, https://doi.org/10.1029/97GL03094.
Takaya, K., and H. Nakamura, 2001: A formulation of a phase-independent wave-activity flux for stationary and migratory quasigeostrophic eddies on a zonally varying basic flow. J. Atmos. Sci., 58, 608–627, https://doi.org/10.1175/1520-0469(2001)058<0608:AFOAPI>2.0.CO;2.
Tapp, R. G., and S. L. Barrell, 1984: The north-west Australian cloud band: Climatology, characteristics, and factors associated with development. J. Climate, 4, 411–424, https://doi.org/10.1002/joc.3370040406.
Telcik, N., and C. Pattiaratchi, 2014: Influence of northwest cloudbands on southwest Australian rainfall. J. Climatol., 2014, 671394, https://doi.org/10.1155/2014/671394.
Tozer, C. R., J. S. Risbey, D. P. Monselesan, D. T. Squire, M. A. Chamberlain, R. J. Matear, and T. Ziehn, 2020: Assessing the representation of Australian regional climate extremes and their associated atmospheric circulation in climate models. J. Climate, 33, 1227–1245, https://doi.org/10.1175/JCLI-D-19-0287.1.
Ummenhofer, C. C., M. H. England, P. C. McIntosh, G. A. Meyers, M. J. Pook, J. S. Risbey, A. S. Gupta, and A. S. Taschetto, 2009: What causes southeast Australia’s worst droughts? Geophys. Res. Lett., 36, L04706, https://doi.org/10.1029/2008GL036801.
Verdon-Kidd, D. C., and A. S. Kiem, 2009: Nature and causes of protracted droughts in southeast Australia: Comparison between the Federation, WWII, and Big Dry droughts. Geophys. Res. Lett., 36, L22707, https://doi.org/10.1029/2009GL041067.
Webster, P. J., V. O. Magan̄a, T. N. Palmer, J. Shukla, R. A. Tomas, M. Yanai, and T. Yasunari, 1998: Monsoons: Processes, predictability, and the prospects for prediction. J. Geophys. Res., 103, 14 451–14 510, https://doi.org/10.1029/97JC02719.
Wright, W. J., 1997: Tropical-extratropical cloudbands and Australian rainfall: I. Climatology. Int. J. Climatol., 17, 807–829, https://doi.org/10.1002/(SICI)1097-0088(19970630)17:8<807::AID-JOC162>3.0.CO;2-J.