Abstract

Macquarie Island (54.50°S, 158.94°E) is an isolated island with modest orography in the midst of the Southern Ocean with precipitation records dating back to 1948. These records (referred to as MAC) are of particular interest because of the relatively large biases in the energy and water budgets commonly found in climate simulations and reanalysis products over the region. A basic climatology of the surface precipitation P is presented and compared with the ERA-Interim (ERA-I) reanalysis. The annual ERA-I precipitation (953 mm) is found to underestimate the annual MAC precipitation (1023 mm) by 6.8% from 1979 to 2011. The frequency of 3-h surface precipitation at MAC is 36.4% from 2003 to 2011. Light precipitation (0.066 ≤ P < 0.5 mm h−1) dominates this dataset (29.7%), and heavy precipitation (P ≥ 1.5 mm h−1) is rare (1.1%). Drizzle (0 < P < 0.066 mm h−1) is commonly produced by ERA-I (43.9%) but is weaker than the detectable threshold of MAC. Warm rain intensity and frequency from CloudSat products were compared with those from MAC. These CloudSat products also recorded considerable drizzle (16%–30%) but were not significantly different from MAC when P ≥ 0.5 mm h−1. Heavy precipitation events were, in general, more commonly associated with fronts and cyclonic lows. Some heavy precipitation events were found to arise from weaker fronts and lows that were not adequately represented in the reanalysis products. Yet other heavy precipitation events were observed at points/times not associated with either fronts or cyclonic lows. Two case studies are employed to further examine this finding.

1. Introduction

The atmospheric environment over the Southern Ocean (SO) is unique: the lack of terrestrial and anthropogenic aerosols creates a pristine environment with few cloud condensation nuclei (Yum and Hudson 2004; Gras 1995). Strong winds produce large waves that, when coupled together, generate large concentrations of sea spray (Murphy et al. 1998). Recent satellite observations of the cloud-top thermodynamic phase suggest that vast fields of clouds composed of supercooled liquid water (SLW) are dominant over the region (Hu et al. 2010; Morrison et al. 2011; Huang et al. 2012b) and are considerably different from those over the North Atlantic Ocean (Huang et al. 2015). Limited in situ cloud observations have found that SLW can exist throughout the entire depth of these clouds, which are often hundreds of meters thick (Chubb et al. 2013; Morrison et al. 2010; Ryan and King 1997; Mossop et al. 1970).

Chubb et al. (2013) further observed that the precipitation under these SLW clouds (cloud-top temperature down to −22°C) was of various thermodynamic phases (glaciated, mixed phase, or even entirely supercooled liquid), highlighting our very limited understanding of the nature of precipitation over the SO. Yet such an understanding is necessary to close both the water and energy budgets over this region that covers 15% of Earth’s surface. Trenberth and Fasullo (2010) detailed relatively large biases in the radiative budget over the SO in both reanalysis products and climate models. Further, large biases have been found in the simulation of precipitation over the SO, with an overestimate of drizzle and underestimate of intense precipitation when compared with satellite observations (e.g., Franklin et al. 2013; Catto et al. 2013).

Among current spaceborne sensors, the CloudSat Cloud Profiling Radar (CPR; Stephens et al. 2002) on board the A-Train satellite constellation has been the most sensitive sensor for detecting light rain and drizzle. Berg et al. (2010) reported that a significant amount of light rainfall and drizzle over subtropical and high-latitude oceans was missed by the Tropical Rainfall Measuring Mission but captured by CloudSat. Ellis et al. (2009) employed a CPR precipitation product (2C-COLUMN-PRECIP; Haynes et al. 2009) to quantify the common occurrence of precipitation over the global oceans; a peak in the frequency of precipitation occurrence was observed between 50° and 60°S. Further, they highlighted that at such high latitudes much of the precipitation was actually classified as snow/ice or mixed phase. Stephens et al. (2010) compared the CPR liquid precipitation product with that from five different global numerical models and found that the time-integrated accumulation was largely consistent for the midlatitudes oceans. The frequency of precipitation from these models, however, is approximately twice that from CPR, with the intensity being correspondingly weaker. Mitrescu et al. (2010) details a second CloudSat precipitation product (2C-RAIN-PROFILE), which has a specific focus on quantifying the intensity of light precipitation. An initial climatology found that “very light” precipitation P (0 < P < 1 mm h−1) was dominant over the Southern Ocean, commonly with a frequency in excess of 15%.

Macquarie Island (54.50°S, 158.94°E) is an isolated island in the midst of the Southern Ocean (Fig. 1) and houses an Australian Bureau of Meteorology (ABoM) weather observation station, which has been in operation since 1948 and is maintained by the Australian Antarctic Division. In recent years these observations (referred to hereinafter as MAC) have become quite valuable in helping develop an understanding of the meteorological conditions over the SO. For example, Hande et al. (2012b) analyzed the thermodynamic structure of the routine upper-air soundings to quantify the strong wind shear in the boundary layer over the SO and a corresponding bias in the ERA-Interim (ERA-I) reanalysis. Adams (2009) examined the trends in the surface observations highlighting a 35% increase in the annual precipitation over a 38-yr period from 1971 to 2008. Such strong trends were not evident in the ERA-40 reanalysis. Jovanovic et al. (2012) reported similar trends in precipitation for the station. Adams (2009) went further to highlight a modest increase in the winds over Macquarie Island over this time frame and a corresponding drying of the boundary layer air. Hande et al. (2012a) similarly identified a statistically significant increase in the winds over Macquarie Island of 2.99 cm s−1 yr−1 for 1991–2011. Over the same period of time the ERA-I reanalysis recorded a decrease of 2.21 cm s−1 yr−1. The upper-air soundings also suggest that both CloudSat and a merged radar–lidar product underestimate the fraction of boundary layer clouds below 750 m (Huang et al. 2012a) due to near-surface clutter.

Fig. 1.

Map showing the location of Macquarie Island over the Southern Ocean and the five A-Train segments with 2.5° × 2.5° coverage. The inset shows the terrain of Macquarie Island together with the location of the ABoM station.

Fig. 1.

Map showing the location of Macquarie Island over the Southern Ocean and the five A-Train segments with 2.5° × 2.5° coverage. The inset shows the terrain of Macquarie Island together with the location of the ABoM station.

The primary aim of this paper is to explore the frequency and intensity of the MAC precipitation observations and their relationship with synoptic systems. These observations are directly compared with CloudSat and ERA-I precipitation products that are relied upon for an understanding and modeling of the SO dynamics, clouds, and precipitation. We present two case studies employing broader A-Train observations to further explore the nature of precipitation over the SO and how these various products represent it.

2. Methodology

a. Macquarie Island surface observations

Located in the midst of the Southern Ocean, Macquarie Island is roughly 34 km long from north to south and at most 5.5 km wide. The ABoM station (station 300004) is located far to the north along a narrow isthmus (Fig. 1). The base height of the station is only 6 m above sea level. Standard surface observations are recorded along with twice-daily upper-air soundings. Jovanovic et al. (2012) detail the metadata of these records and the quality control measures. For example, temperatures at the station were recorded by thermometers and regularly checked against the thermograph records; rain rates were measured by a rain gauge and regularly compared with the pluviograph records. The precipitation records were tested to be homogeneous by using the total cloud amount and sea level pressure, as well as RClimDex software (Zhang and Yang 2004). Although warm rain is dominant, glaciated precipitation is not uncommon (Jovanovic et al. 2012). The combination of glaciated precipitation (snow–ice) and high winds has the potential to lead to an underreporting of precipitation, particularly during the cold seasons.

The meteorology over the SO is dominated throughout the year by the circumpolar storm track (e.g., Simmonds and Keay 2000). The average winds, accordingly, are strongly westerly and have a high relative humidity in the boundary layer (Hande et al. 2012b). Given these strong winds and the limited dimensions of the island, it is not surprising that no diurnal cycle is evident in the surface precipitation. These strong, humid winds do, however, have the potential to create a local orographic effect on precipitation. Macquarie Island, however, is also modest with respect to orography; the peak elevation (~410 m) is located ~25 km to the south of the observation station and is rarely directly upwind of the site. A ridge to the west-southwest of the station of height ~200 m has a greater potential to be an upwind barrier as it is less than 3 km away. Wind roses compiled from the upper-air soundings suggest that this ridge does influence the local wind heading for heights below approximately 250 m (Fig. 2a). The observed winds at 900 hPa are comparable to those in the ERA-I reanalysis, while the observed winds at 975 hPa display a shift to the northwest. Below this level, winds with an origin of 235°–270° heading are less frequently observed while those with an origin of 270°–315° are correspondingly more often.

Fig. 2.

Wind roses and precipitation roses over Macquarie Island. (a) Wind roses from (left) sounding and (right) ERA-I data on different pressure levels from 900 to 1000 hPa. (b) Precipitation roses from (left) MAC and (right) ERA-I records with the wind at 900 hPa from ERA-I.

Fig. 2.

Wind roses and precipitation roses over Macquarie Island. (a) Wind roses from (left) sounding and (right) ERA-I data on different pressure levels from 900 to 1000 hPa. (b) Precipitation roses from (left) MAC and (right) ERA-I records with the wind at 900 hPa from ERA-I.

The minimum detected precipitation at MAC is 0.2 mm for the hourly records. Amounts below this are discarded. Accordingly, at this time scale “no precipitation” at MAC actually defines a precipitation rate of 0 ≤ P < 0.2 mm h−1. We define “light” precipitation as 0.2 ≤ P < 0.5 mm h−1, “moderate” precipitation as 0.5 ≤ P < 1.5 mm h−1, and “heavy” precipitation as P ≥ 1.5 mm h−1), which represents the 1% most extreme events observed at MAC. Because other precipitation products over Macquarie Island, such as those from a reanalysis product, are not limited by this 0.2 mm h−1 detection threshold, we define “drizzle” as a positive quantity of precipitation below MAC detection with 0 < P < 0.2 mm h−1.

For comparison purposes (as described in section 2c), longer time averaging periods (3 and 6 h) are also examined. The minimum MAC detection rate drops accordingly (0.066 and 0.033 mm h−1, respectively). The definitions of drizzle and light precipitation also shift accordingly. Specifically when working with 3-h observations, drizzle is defined as 0 ≤ P < 0.066 mm h−1 and light precipitation is defined as 0.066 ≤ P < 0.5 mm h−1. The definitions of moderate and heavy precipitation are constant throughout this paper.

b. ERA-Interim products

ERA-I is the latest global atmospheric reanalysis product developed by the European Centre for Medium-Range Weather Forecasts (ECMWF), covering the period from 1979 onward (Dee et al. 2011). The spatial resolution of the dataset is 0.75°× 0.75° on 60 vertical levels. The gridded data include 3-hourly surface fields and 6-hourly upper-air atmospheric fields. The surface precipitation is the sum of two components (stratiform and convective), which are computed separately in the model. The temporal resolution of the ERA-I precipitation data allows for a direct comparison with the hourly MAC surface precipitation records that have been summed to either a 3- or 6-h time interval.

c. CloudSat precipitation products

The CloudSat precipitation data employed in this study are the 2C‐COLUMN‐PRECIP (hereinafter PC; Haynes et al. 2009) and 2C-RAIN-PROFILE (hereinafter RP; Mitrescu et al. 2010; Lebsock and L’Ecuyer 2011) products. The PC product provides the presence, and often intensity, of precipitation derived from estimates of the path-integrated attenuation (PIA) of the radar signal, which is determined using an empirical relationship between the clear-sky surface backscattering cross section, surface wind speed, sea surface temperature, and atmospheric temperature and moisture profiles over the oceans. Surface rain rate is retrieved by assuming the invariability of vertical rainfall profiles. The more recently developed RP product incorporates the vertical variability of rainwater and drop-size distributions, in addition to PIA and the observed reflectivity profiles. In this study we have examined all segments (2007–10) from the five tracks nearest the MAC station (Fig. 1). In total, 417 segments are analyzed. (Note that a battery anomaly within the satellite caused missing data for 40 days in December 2009 and January 2010.)

d. Scale and sampling aspects

The temporal and spatial geometries of these various products are vastly different, confounding any direct comparison. The CPR measurements for a single column are approximately 0.16 s and about 1.4 km wide. The CPR orbit segments analyzed here were measured in tens of seconds thus producing a near-instantaneous, one-dimensional cross section. Precipitation rates used in this study are averages of the precipitation columns along the segments. The ERA-I precipitation product covers a time period of 3 h and is two-dimensional (0.75° × 0.75°). Stephens et al. (2010) discussed in detail the difficulty in directly comparing CloudSat observations with those from a reanalysis product because of differences in temporal and spatial geometry. A-Train segments of both onefold (the model grid length) and threefold (3 times the model grid length) were analyzed to explore the sensitivity of the results to the scaling factor. One can conceptually employ a mean wind speed over the station to turn the surface observations into a one-dimensional spatial sample. The mean surface wind speed of Macquarie Island was ~10 m s−1 from 2003 to 2011, while the wind speed at 900 hPa was ~16 m s−1. A 3-h accumulation from the surface MAC observations covers roughly 140 km, which is on the same order of magnitude as these 0.75° and 2.5° A-Train segments. The 3-h MAC records are employed when comparisons are made with the CloudSat products.

The majority of the low-level clouds over this region of the SO have temperatures between 0° and −20°C (Huang et al. 2012a,b). In this temperature range, the thermodynamic phase of both clouds and precipitation may be ambiguous, which makes it difficult to interpret the melting rate of any ice/snow and its density. While this does not affect the analysis of the frequency of precipitation, it does affect the intensity. As a result, our analysis of the intensity using the PC and RP products is limited to when “warm” (liquid) rain is designated for the lowest layers. When the precipitation is deemed to be either mixed phase or glaciated, no precipitation rate is retrieved. Tables 1 and 2 detail the frequencies with which mixed-phase and glaciated radar columns were encountered, overall and by segments of 2.5° length (267 radar columns). Note that the 2C-SNOW-PROFILE (Mitrescu et al. 2010) dataset does provide an explicit measure of the intensity of frozen precipitation, but is not discussed here because of the very limited number of observations available.

Table 1.

The frequency of the thermodynamic state of precipitation from the radar columns (including certain, probable, and possible) of PC products over the 4 yr from 2007 to 2010.

The frequency of the thermodynamic state of precipitation from the radar columns (including certain, probable, and possible) of PC products over the 4 yr from 2007 to 2010.
The frequency of the thermodynamic state of precipitation from the radar columns (including certain, probable, and possible) of PC products over the 4 yr from 2007 to 2010.
Table 2.

The frequency of the thermodynamic state of the precipitation within a 267-radar-columns segment (2.5°). A total of 417 segments are considered over the 4 yr from 2007 to 2010. Warm rain represents the segments consisting of the pixels flagged as both warm rain and clear; glaciation represents segments consisting of the radar columns flagged as both glaciation and clear; and the “other” category includes the mixed phase and segments consisting of the radar columns flagged as both warm rain and glaciation.

The frequency of the thermodynamic state of the precipitation within a 267-radar-columns segment (2.5°). A total of 417 segments are considered over the 4 yr from 2007 to 2010. Warm rain represents the segments consisting of the pixels flagged as both warm rain and clear; glaciation represents segments consisting of the radar columns flagged as both glaciation and clear; and the “other” category includes the mixed phase and segments consisting of the radar columns flagged as both warm rain and glaciation.
The frequency of the thermodynamic state of the precipitation within a 267-radar-columns segment (2.5°). A total of 417 segments are considered over the 4 yr from 2007 to 2010. Warm rain represents the segments consisting of the pixels flagged as both warm rain and clear; glaciation represents segments consisting of the radar columns flagged as both glaciation and clear; and the “other” category includes the mixed phase and segments consisting of the radar columns flagged as both warm rain and glaciation.

Over Macquarie Island, mixed-phase precipitation is more frequently recorded than glaciated (Tables 1 and 2). “Glaciation segments” only occur 5.45% of the time in winter, so including the 2C-SNOW-PROFILE would not eliminate any seasonal bias. Although more than 60% of all individual radar columns are “clear” of any precipitation in each season (Table 1), segments that contain at least one glaciation or mixed-phase radar column among the 267 radar columns are more commonly observed than the clear-only segments, especially in winter (Table 2). Radar columns identified as “certain” (unattenuated near-surface reflectivity of 0 dBZ or higher), “probable” (unattenuated near-surface reflectivity between −7.5 and 0 dBZ), and “possible” (unattenuated near-surface reflectivity < −7.5 dBZ) have been employed here (Haynes et al. 2009).

While it is not possible to calculate the annual mean precipitation rate from CloudSat because of the mixed-phase bias, it is still readily possible to record and analyze the frequency of precipitation (e.g., Ellis et al. 2009). In addition, the precipitation rate of the warm rain is available for analysis.

3. Analysis of precipitation

a. Frequency and intensity

Using the 900-hPa winds and surface precipitation, a precipitation rose (Fig. 2b) reveals that the vast majority of the surface precipitation (MAC) occurs at relatively weak precipitation rates. Not surprisingly, precipitation arrives predominantly from the west. At this coarse 6-h time scale, light precipitation is present 43.8% of the time, moderate precipitation is present 5.8% of the time, and heavy precipitation is present 0.7% of the time. No precipitation was recorded for the remaining 49.7% of the 6-h time blocks. To first order, this is comparable to the statistics from ERA-I, which shows 48.1%, 4.3%, and 0.2% in the light, moderate, and heavy precipitation categories, respectively. ERA-I also produces a large amount (35.4%) of drizzle (0 < P < 0.033 mm h−1), which is too weak to be recorded at MAC.

If the frequency of precipitation in each month is treated as an independent sample, then we have 108 samples in the 9 yr examined. A paired t test shows the difference between MAC and ERA-I is significant for both light precipitation and heavy precipitation at a 95% confidence level. This suggests that the reanalysis product slightly overpredicts the frequency of light precipitation and underpredicts the frequency of heavy precipitation at the 6-h time scale. Figure 2b also shows that the heavy precipitation comes preferentially from the northwest, consistent with the predominant position of cyclonic lows approaching Macquarie Island (Simmonds and Keay 2000).

Instead of displaying a sinusoidal annual cycle, as might be expected, the historic monthly precipitation (1979–2011) at MAC is relatively flat through much of the year (~80 mm month−1) with a single peak (~100 mm month−1) in early autumn (March and April) (Fig. 3). The ERA-I average monthly precipitation (1979–2011) similarly peaks in the autumn with minima in both July and December with a weak increase in between. Over this common period the average annual precipitation for the surface observations is 1023 mm yr−1, which is slightly greater than that from ERA-I (953 mm yr−1). Focusing on the early surface observations (1948–78), the shape of the annual cycle changes little but the annual average precipitation decreases to 971 mm yr−1, consistent with the analysis of Adams (2009). The precipitation accumulation decreases in every month, but only the decrease for August is significant.

Fig. 3.

Average monthly precipitation (MAC 1948–78, MAC 1979–2011, and ERA-I 1979–2011). Error bars show the standard errors calculated from the accumulation in each month.

Fig. 3.

Average monthly precipitation (MAC 1948–78, MAC 1979–2011, and ERA-I 1979–2011). Error bars show the standard errors calculated from the accumulation in each month.

Limiting the analysis to the 4-yr period 2007–10, a direct temporal comparison of the frequency of precipitation can be made for the 3-h MAC observations, ERA-I, and the CPR radar columns. The overall frequency of the MAC precipitation is 36.8%. If each month is again treated as an independent sample, the 95% confidence interval for this value is 35.1%–38.5%. The frequency of the 3-h ERA-I precipitation is 31.9% with 95% confidence interval of 29.8%–33.9%. Here, we have enforced a minimum threshold for ERA-I precipitation of 0.066 mm h−1 to be consistent with the surface observations. The mean precipitation rate from MAC (0.32 mm h−1) is larger than that from ERA-I (0.29 mm h−1), and the monthly variability (0.31, 0.33 mm h−1) is also underestimated by ERA-I (0.29, 0.30 mm h−1).

The frequency for the CPR PC rain (certain, probable, and possible radar columns) is 30.5% with a 95% confidence interval of 27.1%–33.9% but reduces to 21.5% with a 95% confidence interval of 18.4%–24.6% if the possible column is omitted. The mean rain rate of certain radar columns only is 1.24 mm h−1 with a 95% confidence interval of 1.19–1.29 mm h−1. This value is larger than the segment mean rain rate because it excludes the clear radar columns in rain segments, which are common. The mean rain rates of probable and possible radar columns are 0.28 mm h−1 with 95% confidence interval of 0.27–0.29 mm h−1 and 0.13 mm h−1 with 95% confidence interval of 0.12–0.14 mm h−1, respectively. As the CPR RP product only employs “certain” radar columns to avoid the drizzle that does not potentially reach the surface (Lebsock and L’Ecuyer 2011), its precipitation frequency is quite low at 16.3%.

A probability distribution function (PDF) of the precipitation rate can be calculated for the MAC observations (3-h time intervals, 2003–11) and the 3-h ERA-I precipitation product (Fig. 4a). For the MAC observations, “no detectable” precipitation (0 ≤ P < 0.066 mm h−1) is encountered 63.6% of the time, and the frequency falls off quickly as the precipitation rate increases from light (29.7%) to moderate (7.7%) to heavy (1.1%). ERA-I produces frequencies of 23.6%, 43.8%, 27.6%, 4.6%, and 0.3% for “no precipitation” and drizzle, light, moderate, and heavy precipitation, respectively. The frequency of heavy precipitation from ERA-I is less than that from MAC. This can be a systematic difference between the data from a single surface station and the averaging in a model grid, because 3 h is shorter than a storm’s lifetime. But Pfeifroth et al. (2013) compared monthly precipitation data from ERA-I with those from rain gauges in the Pacific Ocean and found that ERA-I systematically underestimates high precipitation.

Fig. 4.

PDF of precipitation intensity. (a) PDF of all precipitation intensity for 3-h MAC and ERA-I from 2003 to 2011. (b) PDF of warm-rain-only intensity (limiting CloudSat to 176 warm rain or clear cases and the corresponding 3-h MAC cases). PC1 and RP1 represent 0.75°-latitude segments in the CloudSat products, and PC3 and RP3 represent 2.5°-latitude segments.

Fig. 4.

PDF of precipitation intensity. (a) PDF of all precipitation intensity for 3-h MAC and ERA-I from 2003 to 2011. (b) PDF of warm-rain-only intensity (limiting CloudSat to 176 warm rain or clear cases and the corresponding 3-h MAC cases). PC1 and RP1 represent 0.75°-latitude segments in the CloudSat products, and PC3 and RP3 represent 2.5°-latitude segments.

Given that both the PC and RP products do not produce the rain rate when mixed-phase or ice precipitation radar columns are encountered, it is not possible to directly extend the comparison to these CPR products. It is possible, however, to make such a PDF for a very limited selection of segments when neither CloudSat products reported mixed-phase or ice precipitation columns. Only 176 out of the 417 threefold segments meet this criterion; these reduced datasets are hereinafter referred to as RP1, RP3, PC1, and PC3 as both the one- and threefold length segments are analyzed. These CloudSat PDFs may then be compared against those from the corresponding 176 3-h MAC observations (Fig. 4b). Note that such calculations introduce a strong seasonal bias, as more snow/ice flags were recorded during the winter and spring seasons (Tables 1 and 2). Further note that the rain rates have been set to 0 for any possible radar columns for the PC1 and PC3 datasets to be consistent with the studies of Ellis et al. (2009) and Stephens et al. (2010). The MAC observations produce frequencies of 67.2% for no detectable precipitation, which is similar to no precipitation (P = 0 mm h−1) RP1 (66.5%) but higher than those of RP3, PC1, and PC3.

The CPR RP very likely underestimates the occurrence of light precipitation (defined as 0.066–0.5 mm h−1) but compares reasonably well (given the inherent space–time difficulties) for higher precipitation rates. This is perhaps not completely surprising since CPR RP derives precipitation rates only for radar columns where precipitation is “certain.” The CPR PC may also be underestimating light precipitation, but any underestimate is small (relative to that of the CPR RP). Given the ambiguity of the MAC soundings at intensities below 0.066 mm h−1, it is not worthwhile to label any one of the CloudSat products as superior to the others; rather, the limited surface observations would generally provide some confidence in these satellite observations for this region of the Southern Ocean.

The individual matchups between MAC and the CloudSat products at a smaller time and spatial scale were further investigated using simple contingency tables at various intensity thresholds. Mean rain rates from segments of only 50-km length were retrieved from the nearest three orbits to MAC, and were matched with the corresponding 1-h MAC observations. A total of 141 “warm rain only” segments are available for this analysis. At this scale, both the PC (Table 3) and RP (Table 4) products show approximately 90% agreement (hits + misses) with MAC in all the comparisons using different minimum detection thresholds (0.2, 0.5, and 1.0 mm h−1). MAC records slightly more precipitation events than either CloudSat product. Since the PC product includes the probable precipitation radar columns, it matches MAC in frequency better than does the RP product for light precipitation.

Table 3.

Contingency table showing the comparison of precipitation occurrences using the CloudSat PC product and surface observations (MAC) for the 141 clear and warm rain cases in the 50-km segments.

Contingency table showing the comparison of precipitation occurrences using the CloudSat PC product and surface observations (MAC) for the 141 clear and warm rain cases in the 50-km segments.
Contingency table showing the comparison of precipitation occurrences using the CloudSat PC product and surface observations (MAC) for the 141 clear and warm rain cases in the 50-km segments.
Table 4.

Contingency table showing the comparison of precipitation occurrences using the CloudSat RP product and surface observations (MAC) for the 141 clear and warm rain cases in the 50-km segments.

Contingency table showing the comparison of precipitation occurrences using the CloudSat RP product and surface observations (MAC) for the 141 clear and warm rain cases in the 50-km segments.
Contingency table showing the comparison of precipitation occurrences using the CloudSat RP product and surface observations (MAC) for the 141 clear and warm rain cases in the 50-km segments.

A scatterplot (Fig. 5) also shows these individual matchups between MAC and CloudSat. Again, the RP product missed more light precipitation recorded by MAC (0.2, 0.8 mm h−1) than the PC product. Paired t tests find that the difference between PC and MAC is not significant at the 95% confidence level, but the difference between RP and MAC is significant.

Fig. 5.

Scatterplot of rain rates from MAC and two CloudSat products (PC and RP) for 141 clear and warm rain cases in 50-km segments.

Fig. 5.

Scatterplot of rain rates from MAC and two CloudSat products (PC and RP) for 141 clear and warm rain cases in 50-km segments.

b. Precipitation and synoptic weather systems

As the synoptic meteorology over the SO is dominated by the circumpolar storm track, it is of interest to identify how much precipitation occurs as a result of frontal activity and how much is not immediately associated with fronts (e.g., Catto et al. 2013). Adams (2009) partially attributed the increasing trend in rainfall over Macquarie Island to an increase in cyclonic activity.

There are a variety of means of identifying fronts and midlatitude cyclones, each with its own strengths and weaknesses; we consider four such methods. Historically, the ABoM produces hand-drawn daily forecasts of major cyclones and associated cold fronts over the SO, which are primarily based on coarse global numerical weather prediction output and satellite observations (hereinafter referred to as the ABoM method). Such forecasts commonly capture only prominent synoptic-scale events, with a potential for missing smaller transient mesoscale events. Further, it has been a common operational practice of the ABoM to not forecast warm fronts, as the vast majority of such events occur at latitudes too high to affect the Australian mainland.

Berry et al. (2011), whose method is hereinafter referred to as BRJ, produced a climatology of fronts from the global reanalysis product ERA-40. Fronts were defined by the gradient in the horizontal wet-bulb potential temperature at 850 hPa. This method identifies cold, warm, and quasi-stationary fronts, with the overall peak frequency in fronts found along the storm tracks over the SO, North Pacific, and North Atlantic.

Adams (2009) used the local surface pressure p measurements to identify when low pressure systems were passing over Macquarie Island (hereinafter referred to as the ASP method). Specifically, a positive second derivative of pressure with respect to time d2p/dt2 was used to identify times when the pressure was concave upward. No distinction was made between warm, cold, and stationary fronts and low pressure systems. When a preselected threshold was passed (e.g., d2p/dt2 > 0.2 hPa h−2), a “cyclonic activity event” was defined. The number of events identified by the ASP definition was found to increase from roughly 10 per month in 1970 to 15 per month in 2008 (Adams 2009).

We have modified the ASP method by further incorporating the surface observations of wind direction and temperature in 3-h time series (d2p/dt2 > 0.1 hPa h−2, dT/dt < −0.2°C h−1, and ddir/dt < −5° h−1; hereinafter referred to as the MSP method.) These thresholds are induced from two training periods during the winter of 2002 and summer of 2003.

The decrease in pressure may also be used to identify cyclonic lows that may pass directly overhead (dp/dt < −8 hPa h−1).

For any of these methods, MAC observations of precipitation recorded in the 6 h before or after a front are regarded as being associated with that front (or cyclonic low). Also, even though the CPR product is able to distinguish between frontal and nonfrontal rainfall, it is not employed because of the limitations of the precipitation rate and the limited sample size. Given the coarse temporal resolution in the production of the daily mean sea level pressure (MSLP) charts, it is not practical to make such a calculation using the ABoM definition of fronts. They are, however, included in the initial discussion to better appreciate the occurrence of fronts over Macquarie Island.

These various definitions of fronts and cyclonic activity commonly identify strong events, but are not always consistent during other periods, as illustrated by a 2-month sample (July–August 2002) (Fig. 6a). During these two months, ABoM identified 26 fronts, which is relatively consistent with BRJ (24 fronts), though not necessarily the same front or with the same timing. Both of these methods rely on an analysis or reanalysis product. ASP (30 fronts) and MSP (29 fronts) show few differences here, as they both primarily use time series of the surface pressure. The corresponding surface observations (pressure, wind direction, temperature, and precipitation) highlight the variability of the meteorological conditions along the storm track (Fig. 6c).

Fig. 6.

Two-month time series (July–August 2002) of the identified (a) fronts and (b) lows using different definitions (ABoM, ASP, MSP, and BRJ), with (c) the corresponding time series of surface precipitation and thermodynamic variables (p, dir, and T).

Fig. 6.

Two-month time series (July–August 2002) of the identified (a) fronts and (b) lows using different definitions (ABoM, ASP, MSP, and BRJ), with (c) the corresponding time series of surface precipitation and thermodynamic variables (p, dir, and T).

Employing the MAC observations alone, the fraction of the precipitation associated with “cyclonic activity” can be calculated using either the ASP or MSP definitions (Fig. 7). The ASP definition finds that roughly half of the light-precipitation events occur during periods of cyclonic activity and that number increases steadily to over 60% for heavy precipitation events. The MSP definition finds considerably less precipitation to be associated with frontal passages at all intensities. Surprisingly, the amount of precipitation events associated with cyclones, as opposed to fronts, remains steady even as the intensity of the precipitation is increased.

Fig. 7.

Rainfall decomposition into different intensities with (a) ASP definitions (identifying fronts with a threshold of the second derivative of surface pressure) and (b) MSP definitions (identifying fronts with a threshold of the second derivative of surface pressure and the first derivative of wind direction and temperature) using precipitation data from MAC from 2003 to 2011.

Fig. 7.

Rainfall decomposition into different intensities with (a) ASP definitions (identifying fronts with a threshold of the second derivative of surface pressure) and (b) MSP definitions (identifying fronts with a threshold of the second derivative of surface pressure and the first derivative of wind direction and temperature) using precipitation data from MAC from 2003 to 2011.

The impact of the time window (a time interval around the frontal passage time) on the percentage of the frontal precipitation is shown in Fig. 8. ASP identifies more frontal precipitation than MSP at the same window time. Both of their slopes decrease with the window time, which implies that there is more precipitation near the defined frontal time. These two lines become roughly linear when the window time is larger than 12 h, which implies that the effect of the local front is no longer evident and only the duration of the window defines the amount of precipitation. Naud et al. (2012) found precipitation associated with the Southern Hemisphere extratropical cyclones tends to fall within a 10° range across the front. If this distance is divided by the mean wind speed of 16 m s−1 at 900 hPa, the time is also approximately 12 h, consistent with the time window here.

Fig. 8.

The percentage of the precipitation in the frontal window time using the ASP and MSP methods.

Fig. 8.

The percentage of the precipitation in the frontal window time using the ASP and MSP methods.

The BRJ classification may be used with any reanalysis product; in this work we employ the ERA-I reanalysis for consistency. Any surface precipitation record can be decomposed according to this classification. In Fig. 9, the precipitation decomposition is constructed for both the ERA-I and MAC precipitation results (Fig. 7). These decompositions are consistent with ASP and MSP decompositions. At weak precipitation rates, the “other” class is dominant. As the precipitation rate increases, the precipitation event is more likely to be associated with a cold front. This relationship between precipitation rate and the strength of a frontal event is stronger when the ERA-I precipitation is used.

Fig. 9.

Rainfall decomposition into different intensities with the BRJ definition using precipitation data from (a) MAC and (b) ERA-I from 2003 to 2011.

Fig. 9.

Rainfall decomposition into different intensities with the BRJ definition using precipitation data from (a) MAC and (b) ERA-I from 2003 to 2011.

4. Case studies and discussion

Figure 6 highlighted that there is notable inconsistency in identifying fronts with respect to the various algorithms, especially for relatively weak fronts. An example of a weak front is explored in the following first case study. Conversely, the different identification algorithms are found to be largely consistent for strong events. Yet even for these heavy precipitation periods, all of the classifications suggest that roughly 20% of the heavy precipitation is not associated with a front or cyclonic activity. A second case study is present for such an other-class event. The two case studies enable an effort to tie together the various precipitation products at Macquarie Island (MAC, ERA-I, RP, and PC) with the meteorological conditions of the Southern Ocean.

As a first example, an eastward-moving low pressure system is examined on 16 March 2008. The MSLP analysis at 0000 UTC (Fig. 10) displays a weak shortwave trough to the west of the island, which was actually not classified as cyclonic activity or a front by any of the algorithms (ASP, MSP, BRJ, or ABoM). The changes in wind direction and temperature are witnessed at about 0000 UTC. Although there is a pressure change, it does not pass the second-derivative threshold. A strong high pressure system is located over the Tasman Sea with a ridge extending across much of the Southern Ocean to the east of Macquarie Island. The 700-hPa analysis (not shown) also indicates that Macquarie Island is located behind the upper-level ridge. The Moderate Resolution Imaging Spectroradiometer (MODIS) image of cloud-top temperature at ~0350 UTC reveals a thick cold (230 K) cloud band approaching Macquarie Island from the west. Macquarie Island is underneath the front edge of the cloud band. The sounding at 2300 UTC 15 March exhibits a shallow unstable layer below 870 hPa, and an increase in moisture in the midtroposphere. The winds swing from northwesterly below 900 hPa to westerly and southwesterly aloft.

Fig. 10.

Case study A over Macquarie Island on 16 Mar 2008 with (a) MSLP, (b) MODIS cloud-top temperature, (c) MAC soundings, (d) CALIOP categorization (here, HOI is horizontally oriented ice), (e) flags from CloudSat, (f) CPR reflectivity, and (g) rain rate from CloudSat.

Fig. 10.

Case study A over Macquarie Island on 16 Mar 2008 with (a) MSLP, (b) MODIS cloud-top temperature, (c) MAC soundings, (d) CALIOP categorization (here, HOI is horizontally oriented ice), (e) flags from CloudSat, (f) CPR reflectivity, and (g) rain rate from CloudSat.

An A-Train segment passes through the leading edge of the thick cloud band. The Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) cloud thermodynamic phase (Hu et al. 2009) and the CPR radar reflectivity depict a thick, high-altitude anvil across most of the segment, with convection being evident between 55° and 56°S. The lidar signal is likely to be fully attenuated in this region. The CALIOP cloud phase suggests that much of the evident cloud field below ~4-km altitude is composed of liquid water. Boundary layer clouds at altitudes below 1 km are observed to both the north and south of the convective region. The radar reflectivity above an altitude of 8 km is relatively weak (Ze < −10 dBZ). The radar is unable to pick up the low-altitude boundary layer clouds because of their proximity to the surface and the resulting surface clutter (Marchand et al. 2009). The ECMWF temperature analysis is overlaid on this image. The freezing level falls from around 3 km at 50°S to below 500 m at 60°S, with the steepest drop occurring through the heavy convection. The low-elevation clouds to the south of this point are, presumably, composed of supercooled liquid water, according to the CALIOP cloud phase.

Looking over the entire segment, the CPR precipitation product shows a relatively modest precipitation event limited to the convective region. Note that no precipitation rate is recorded at the southern boundary of this precipitation event (56°S), where most of the boundary layer temperature has dropped to below 0°C. The CPR precipitation product records precipitation along this segment but has designated this precipitation as being of “mixed phase.”

Over the 6-h interval shown, the average MAC rain rate is 0.57 mm h−1. Over the last 3 h the average rain rate is 0.73 mm h−1. The CPR precipitation products (taken at 0353 UTC) record a moderate event (0.66 mm h−1 from PC and 0.82 mm h−1 from RP) in a 2.5° segment of the granule. The corresponding ERA-I record is 0.46 mm h−1. At this point in time, none of the methods used in this study identifies a frontal passage or cyclonic activity, highlighting the difficulty in defining such events and the speed at which they can develop over the SO.

The second case study examined (1 November 2010) appears to be an exception to the conventional view of heavy precipitation in the midlatitudes, for this precipitation is also not associated with a front. The average CPR PC rain rate at ~0350 UTC is 2.06 mm h−1, which is actually the greatest recorded for a segment that passes closest to Macquarie Island (Fig. 1) during the 4 yr studied (52 segments). The MSLP at 0000 UTC (Fig. 11) indicates that Macquarie Island is located on the outskirts of a strong anticyclone centered near the southern edge of New Zealand. The blocking high pressure system is accompanied by an elongated trough up through the Tasman Sea to the northwest, an approaching front to the far west, and a frontal passage to the far south. This synoptic pattern creates a sharp pressure gradient across Macquarie Island, constricting the flows to form a tropospheric “jet” from the north to Macquarie Island. Different from the first case, the east-located blocking high enables the transport of moisture and heat from the warmer water, facilitating a rapid saturation of the fast-moving, cooling air. The routine sounding at 2300 UTC 23 October (Fig. 11c) shows a deep, moistened atmosphere with the air mass below 700 hPa being completely saturated. The strong, constant northwesterly wind throughout the troposphere suggests the joint influences of the extending upper-level trough and the high.

Fig. 11.

As in Fig. 10, but for case study B on 1 Nov 2010.

Fig. 11.

As in Fig. 10, but for case study B on 1 Nov 2010.

The MODIS image of cloud-top temperature at ~0350 UTC is consistent with the synoptic analysis, suggesting that an anticyclonic flow is dominant at this time. The MODIS cloud-top temperature shows streaks of cold cloud tops (−40°C) within this field. CALIOP, on the other hand, portrays a field of midaltitude clouds composed of SLW. There are patches of high-altitude cirrus near 50° and 58°S, but none near Macquarie Island. The CloudSat radar portrays active convection to a height of ~5 km at this time near 54°S. Convection near 50° and 58°S is evident, too. It is interesting that CloudSat misses much of the cloud near 52°S that is captured by CALIOP. This is consistent with clouds composed of small droplets that are insufficiently reflective for the radar to detect.

The hourly precipitation rate observed at MAC (Table 5) averages to 1.4 mm h−1 over the 3-h period (0200–0400 UTC), or 0.6 mm h−1 if only the last 3 h (0300–0500 UTC) are used. The CloudSat rain rate for the segment is 2.06 mm h−1 from PC and 2.37 mm h−1 from RP, even though the actual rain rate is probably higher because some of the columns reach the maximum retrievable rain rate or the CPR (Haynes et al. 2009). The rain rate from PC is weaker than that from RP at the peak near 54°S, and PC records probable rain at 56°S, but it is rejected by CloudSat RP.

Table 5.

Hourly precipitation records at MAC on 1 Nov 2010 for the 2010 case study.

Hourly precipitation records at MAC on 1 Nov 2010 for the 2010 case study.
Hourly precipitation records at MAC on 1 Nov 2010 for the 2010 case study.

South of 58°S, the boundary layer temperature has decreased to below freezing, which means that neither the PC nor RP rain rates are available. Near Macquarie Island, however, intense rainfall is evident. The corresponding 3-h ERA-I record (at 0600 UTC) is 0.53 mm h−1, which is an exceptionally large value for the ERA-I records. Given the uniqueness of the meteorological situation of this event, it is not surprising that a front (or cyclonic low) was not diagnosed from any of the various methods. It would be interesting to further explore the potential for such an event to be linked to an Antarctic “atmospheric river” (e.g., Tsukernik and Lynch 2013).

These two case studies highlight some of the difficulties associated with the ground validation of the satellite observations as a result of mismatches in the sensitivity and sampling between remote and in situ measurements and the necessity of a rigorous evaluation. The various precipitation products (MAC, PC, RP, and ERA-I) consistently record heavy precipitation for both cases, even though neither event was classified as being associated with a frontal passage or a cyclonic low. This failure to identify a frontal passage is perhaps surprising for the 2008 case study.

5. Conclusions

The several decades of meteorological observations from Macquarie Island (54.50°S, 158.94°E) are a valuable tool in the effort to understand the energy and water budgets over the Southern Ocean. These precipitation records have been analyzed to improve our understanding of the physical nature of the local precipitation and are then compared with the precipitation products of CloudSat (PC and RP) and ERA-Interim, although the different temporal and spatial geometries of these products hinder a robust comparison.

The annual cycle of the long-term precipitation records (1979–2011) displays a peak in March but is relatively flat for the rest of the year. ERA-I also records a peak in March but shows much greater variability throughout the year. Annually averaged, ERA-I underestimates the precipitation as observed at MAC by approximately 6.8%.

From 2003 to 2011, the frequency of the 3-h MAC precipitation is 36.4%. The frequency decreases with an increase in the precipitation intensity, from 29.7% for light precipitation to 1.1% for heavy precipitation. The overall frequency of the ERA-I precipitation is only slightly lower (32.7%) than is found at MAC during these nine years. The light precipitation in ERA-I (27.6%) is also comparable with MAC, but ERA-I produces less heavy precipitation (0.3%) and a considerable amount (43.9%) of drizzle. This drizzle is below the detectable threshold at MAC.

The CloudSat PC and RP products were limited to segments that only included “warm” rain (i.e., mixed-phase and glaciated precipitation was omitted). This filtering led to a seasonal bias in the results and a relatively small pool of 176 segments for the analysis. Following Stephens et al. (2010), segments were considered at onefold (~0.75°) and threefold (~2.5°) the length of the ERA-I grid resolution. These CPR products overestimated the correlated frequency of “no precipitation” (P = 0 mm h−1) and drizzle relative to MAC, especially at the shorter segment length. The frequencies of light precipitation were correspondingly underestimated, particularly for the RP product. Paired t tests found only the threefold PC statistics to be significantly different from that of MAC.

At higher temporal (1 h) and spatial (~0.5°) resolutions, individual matchups between MAC and CloudSat products show approximately 90% agreement in event detections (hits + misses). The PC product shows greater skill for light precipitation than is found for the RP product primarily because it has a higher frequency as the algorithm includes “probable” precipitation radar columns, whereas the RP product only includes “certain” columns.

Fronts or cyclonic activity are associated with light precipitation roughly half of the time and this number increased to roughly 80% during heavy precipitation. It was further noted that the definition of fronts was commonly inconsistent over the Southern Ocean, especially for weaker events. Also a nontrivial fraction (~20%) of the heavy precipitation is not closely associated with frontal systems. Two cases studies have been examined to tie together the synoptic meteorology and the various precipitation observations. The second case study demonstrates that other mechanisms (e.g., shallow convection) can also generate heavy precipitation over the Southern Ocean, which is not associated with fronts or cyclonic lows.

Precipitation over the Southern Ocean is a critical component of Earth’s water and energy cycle. Thus, understanding its characteristics and mechanisms is a vital step toward improving climate predictions. Given that spaceborne sensors remain the only tool for long-term observations over this remote region, there is a pressing need for better validation and characterization of satellite retrieval products. This is particularly true when a new standard of spaceborne precipitation measurement is set with the integration of next-generation products (e.g., Global Precipitation Measurement). Our study herein sheds some light on addressing this underexplored issue by using very basic, local ground-based measurements. Albeit limited, our findings may potentially benefit future dedicated field campaigns with multiple measurement platforms and more comprehensive facilities in place (http://www.atmos.washington.edu/socrates/SOCRATES_white_paper_Final_Sep29_2014.pdf).

Acknowledgments

This research was funded by ARC Linkage Grant LP120100115 and relies on datasets provided by the Australian Bureau of Meteorology. We thank John M. Haynes for the comments on the analysis of CloudSat data. We thank Gareth Berry for help with the identification of fronts and cyclones. We are also indebted to the two anonymous reviewers whose efforts have greatly improved the manuscript throughout.

REFERENCES

REFERENCES
Adams
,
N.
,
2009
:
Climate trends at Macquarie Island and expectations of future climate change in the sub-Antarctic
.
Pap. Proc. Roy. Soc. Tasmania
,
143
,
1
8
.
Berg
,
W.
,
T.
L’Ecuyer
, and
J. M.
Haynes
,
2010
:
The distribution of rainfall over oceans from spaceborne radars
.
J. Climate
,
49
,
535
543
, doi:.
Berry
,
G.
,
M. J.
Reeder
, and
C.
Jakob
,
2011
:
A global climatology of atmospheric fronts
.
Geophys. Res. Lett.
,
38
,
L04809
, doi:.
Catto
,
J. L.
,
C.
Jakob
, and
N.
Nicholls
,
2013
:
A global evaluation of fronts and precipitation in the ACCESS model
.
Aust. Meteor. Oceanogr. J.
,
63
,
191
203
.
Chubb
,
T. H.
,
J. B.
Jensen
,
S. T.
Siems
, and
M. J.
Manton
,
2013
:
In situ observations of supercooled liquid clouds over the Southern Ocean during the HIAPER Pole-to-Pole Observation campaigns
.
Geophys. Res. Lett.
,
40
,
5280
5285
, doi:.
Dee
,
D.
, and Coauthors
,
2011
:
The ERA-Interim reanalysis: Configuration and performance of the data assimilation system
.
Quart. J. Roy. Meteor. Soc.
,
137
,
553
597
, doi:.
Ellis
,
T. D.
,
T.
L’Ecuyer
,
J. M.
Haynes
, and
G. L.
Stephens
,
2009
:
How often does it rain over the global oceans? The perspective from CloudSat
.
Geophys. Res. Lett.
,
36
,
L03815
, doi:.
Franklin
,
C. N.
,
Z.
Sun
,
D.
Bi
,
M.
Dix
,
H.
Yan
, and
A.
Bodas-Salcedo
,
2013
:
Evaluation of clouds in ACCESS using the satellite simulator package COSP: Global, seasonal, and regional cloud properties
.
J. Geophys. Res. Atmos.
,
118
,
732
748
, doi:.
Gras
,
J. L.
,
1995
:
CN, CCN and particle size in Southern Ocean air at Cape Grim
.
Atmos. Res.
,
35
,
233
251
, doi:.
Hande
,
L. B.
,
S. T.
Siems
, and
M. J.
Manton
,
2012a
:
Observed trends in wind speed over the Southern Ocean
.
Geophys. Res. Lett.
,
39
,
L11802
, doi:.
Hande
,
L. B.
,
S. T.
Siems
,
M. J.
Manton
, and
D.
Belusic
,
2012b
:
Observations of wind shear over the Southern Ocean
.
J. Geophys. Res.
,
117
,
D12206
, doi:.
Haynes
,
J. M.
,
T. S.
L’Ecuyer
,
G. L.
Stephens
,
S. D.
Miller
,
C.
Mitrescu
,
N. B.
Wood
, and
S.
Tanelli
,
2009
:
Rainfall retrieval over the ocean with spaceborne W-band radar
.
J. Geophys. Res.
,
114
,
D00A22
, doi:.
Hu
,
Y.
, and Coauthors
,
2009
:
CALIPSO/CALIOP cloud phase discrimination algorithm
.
J. Atmos. Oceanic Technol.
,
26
,
2293
2309
, doi:.
Hu
,
Y.
,
S.
Rodier
,
K. M.
Xu
,
W.
Sun
,
J.
Huang
,
B.
Lin
,
P.
Zhai
, and
D.
Josset
,
2010
:
Occurrence, liquid water content, and fraction of supercooled water clouds from combined CALIOP/IIR/MODIS measurements
.
J. Geophys. Res.
,
115
,
D00H34
, doi:.
Huang
,
Y.
,
S. T.
Siems
,
M. J.
Manton
,
L. B.
Hande
, and
J. M.
Haynes
,
2012a
:
The structure of low-altitude clouds over the Southern Ocean as seen by CloudSat
.
J. Climate
,
25
,
2535
2546
, doi:.
Huang
,
Y.
,
S. T.
Siems
,
M. J.
Manton
,
A.
Protat
, and
J.
Delanoë
,
2012b
:
A study on the low-altitude clouds over the Southern Ocean using the DARDAR-MASK
.
J. Geophys. Res.
,
117
,
D18204
, doi:.
Huang
,
Y.
,
A.
Protat
,
S. T.
Siems
, and
M. J.
Manton
,
2015
:
A-Train observations of maritime midlatitude storm-track cloud systems: Comparing the Southern Ocean against the North Atlantic
.
J. Climate
,
28
,
1920
1939
, doi:.
Jovanovic
,
B.
,
K.
Braganza
,
D.
Collins
, and
D.
Jones
,
2012
:
Climate variations and change evident in high-quality climate data for Australia’s Antarctic and remote island weather stations
.
Aust. Meteorol. Oceanogr. J.
,
62
,
247
261
.
Lebsock
,
M. D.
, and
T. S.
L’Ecuyer
,
2011
:
The retrieval of warm rain from CloudSat
.
J. Geophys. Res.
,
116
,
D20209
, doi:.
Marchand
,
R. T.
,
J.
Haynes
,
G. G.
Mace
,
T.
Ackerman
, and
G.
Stephens
,
2009
:
A comparison of simulated cloud radar output from the multiscale modeling framework global climate model with CloudSat cloud radar observations
.
J. Geophys. Res.
,
114
,
D00A20
, doi:.
Mitrescu
,
C.
,
T.
L’Ecuyer
,
J. M.
Haynes
,
S. D.
Miller
, and
F. J.
Turk
,
2010
:
CloudSat precipitation profiling algorithm—Model description
.
J. Appl. Meteor. Climatol.
,
49
,
991
1003
, doi:.
Morrison
,
A. E.
,
S. T.
Siems
,
M. J.
Manton
, and
A.
Nazarov
,
2010
:
A modeling case study of mixed-phase clouds over the Southern Ocean and Tasmania
.
Mon. Wea. Rev.
,
138
,
839
862
, doi:.
Morrison
,
A. E.
,
S. T.
Siems
, and
M. J.
Manton
,
2011
:
A three-year climatology of cloud-top phase over the Southern Ocean and North Pacific
.
J. Climate
,
24
,
2405
2418
, doi:.
Mossop
,
S. C.
,
A.
Ono
, and
E. R.
Wishart
,
1970
:
Ice particles in maritime clouds near Tasmania
.
Quart. J. Roy. Meteor. Soc.
,
25
,
487
508
, doi:.
Murphy
,
D. M.
, and Coauthors
,
1998
:
Influence of sea-salt on aerosol radiative properties in the Southern Ocean marine boundary layer
.
Nature
,
392
,
62
65
, doi:.
Naud
,
C. M.
,
D. J.
Posselt
, and
S. C.
Van den Heever
,
2012
:
Observational analysis of cloud and precipitation in midlatitude cyclones: Northern versus Southern Hemisphere warm fronts
.
J. Climate
,
25
,
5135
5151
, doi:.
Pfeifroth
,
U.
,
R.
Mueller
, and
B.
Ahrens
,
2013
:
Evaluation of satellite-based and reanalysis precipitation data in the tropical Pacific
.
J. Appl. Meteor. Climatol.
,
52
,
634
644
, doi:.
Ryan
,
B. F.
, and
W. D.
King
,
1997
:
A critical review of the Australian experience in cloud seeding
.
Bull. Amer. Meteor. Soc.
,
78
,
239
254
, doi:.
Simmonds
,
I.
, and
K.
Keay
,
2000
:
Mean Southern Hemisphere extratropical cyclone behavior in the 40-Year NCEP–NCAR Reanalysis
.
J. Climate
,
13
,
873
885
, doi:.
Stephens
,
G. L.
, and Coauthors
,
2002
:
The CloudSat mission and the A-Train
.
Bull. Amer. Meteor. Soc.
,
83
,
1771
1790
, doi:.
Stephens
,
G. L.
, and Coauthors
,
2010
:
Dreary state of precipitation in global models
.
J. Geophys. Res.
,
115
,
D24211
, doi:.
Trenberth
,
K. E.
, and
J. T.
Fasullo
,
2010
:
Simulation of present-day and twenty-first-century energy budgets of the southern oceans
.
J. Climate
,
23
,
440
454
, doi:.
Tsukernik
,
M.
, and
A. H.
Lynch
,
2013
:
Atmospheric meridional moisture flux over the Southern Ocean: A story of the Amundsen Sea
.
J. Climate
,
26
,
8055
8064
, doi:.
Yum
,
S. S.
, and
J. G.
Hudson
,
2004
:
Wintertime/summertime contrasts of cloud condensation nuclei and cloud microphysics over the Southern Ocean
.
J. Geophys. Res.
,
109
,
D06204
, doi:.
Zhang
,
X.
, and
F.
Yang
,
2004
: RClimDex (1) user manual. Environment Canada Climate Research Branch Rep., 23 pp. [Available online at http://etccdi.pacificclimate.org/RClimDex/RClimDexUserManual.doc.]