Clim Dyn DOI 10.1007/s00382-015-2606-1
Spatial and temporal variations in rainfall over Darwin and its vicinity during different large‑scale environments Surendra P. Rauniyar1 · Kevin J. E. Walsh1
Received: 10 July 2014 / Accepted: 10 April 2015 © Springer-Verlag Berlin Heidelberg 2015
Abstract This study analyses the regional variations in rainfall over Darwin and its vicinity due to different largescale circulations during the Australian summer by utilizing the combination of in situ and C-band polarimetric radar rainfall data at hourly resolution. The eight phases of the Madden–Julian oscillation as defined by Wheeler and Hendon (Mon Weather Rev 132(8):1917–1932, 2004) were used as indicators of different large-scale environments. The analysis found that the large-scale forcing starts to build up from phase 4 by the reversal of low- to midlevel easterly winds to moist westerly winds, reaching a maximum in phase 5 and weakening through phases 6–7. During phases 4–6, most of the study domain experiences widespread rainfall, but with distinct spatial and temporal structures. In addition, during these phases, coastal areas near Darwin receive more rainfall in the early morning (0200–0400 LT) due to the spreading or expansion of rainfall from the Beagle Gulf, explaining the occurrence of a secondary diurnal rainfall peak over Darwin. In contrast, local-scale mechanisms (sea breezes) reinvigorate from phase 8, further strengthening through phases 1–3, when low-level easterly winds become established over Darwin producing rainfall predominately over land and island locations during the afternoon. During these phases, below average rainfall is observed over most of the radar domain, except over the Tiwi Islands in phase 2.
* Surendra P. Rauniyar
[email protected] Kevin J. E. Walsh
[email protected] 1
School of Earth Sciences, The University of Melbourne, Victoria 3010, Australia
Keywords Regional variation · Diurnal cycle · MJO · Darwin atmospheric regimes · Land–sea thermal contrast · CPOL radar
1 Introduction Darwin, which lies at the “top end” of Northern Australia (Fig. 1), experiences a wide range of tropical weather during its wet season due to its unique location on the southern edge of the Maritime Continent (Ramage 1968), one of the largest source of latent heating around the globe (Keenan et al. 1989). The weather over Darwin varies significantly in dynamic and thermodynamic properties and generates a wide variety of convection/rainfall regimes (Keenan and Carbone 1992; Drosdowsky 1996; May and Ballinger 2007; Caine et al. 2009; Pope et al. 2009a). Hence, it is believed that analysis of convective regimes over Darwin may be helpful to understand the variations in rainfall across larger areas of tropics (May et al. 2008). Therefore, numerous field experiments have been conducted over Darwin and surrounding regions in the past, such as the Australian Monsoon Experiment (AMEX; Holland et al. 1986), the Island Thunderstorm Experiment (ITEX; Keenan et al. 1989), the Maritime Continental Thunderstorm Experiment (MCTEX; Keenan et al. 2000; Carbone et al. 2000), the Darwin Wave Experiment (DAWEX; Hamilton et al. 2004), and recently the Tropical Warm Pool-International Cloud Experiment (TWP-ICE; May et al. 2008) to examine the characteristics of clouds and rainfall associated with these regimes. Although these field experiments were designed to improve the understanding of certain specific phenomena, together they provide comprehensive longterm observations of meteorological variables (May and Ballinger 2007) and make Darwin an ideal location to investigate the
13
S. P. Rauniyar, K. J. E. Walsh
Fig. 1 Location map of study area showing topography and bathymetrical features in shading. The white filled triangle is the location of C-band polarimetric radar (CPOL) at Gunn Point. The black ring (150 km radius) shows the sampling domain of CPOL, however only the grids within the two concentric red rings (i.e., ranges 20–120 km) are analyzed to reduce sampling uncertainty. The filled squares are the rain gauges operated by the Australian Bureau of Meteorology (BoM) and provide daily accumulated rainfall (mm day−1). The white filled square is the BoM operated rain gauge at Darwin airport whereas the white filled circle is the Atmospheric Radiation Measurement (ARM) Climate Research Facility (ACRF) operated rain gauge. These two stations provide hourly rainfall with an uncertainty of ±0.1 mm h−1
mechanisms behind the spatial and temporal variations in rainfall. The thermodynamic and dynamic properties of convection/rainfall over Darwin and its surrounding regions have been studied generally by separating the wet season into active and build-up/break periods (e.g., Keenan and Carbone 1992; McBride and Frank 1999; May and Ballinger 2007; May et al. 2012). However, various indices have been used to define the active monsoon period and were mainly based on wind and rainfall (Troup 1961; Hendon and Liebmann 1990a); wind only (Drosdowsky 1996); or wind and outgoing longwave radiation (Hung and Yanai 2004). Nevertheless, the principal feature during the active periods in Darwin is the presence of westerly winds of equatorial origin below 700 hPa (May et al. 2008), while low- to midlevel easterly winds of continental origin dominate during the build-up/break periods (Keenan et al. 1989). In general, convection during the active periods tends to be heavy and more widespread (over land and ocean) due to the existence of large-scale ascent and moist mid-levels, however it exhibits a smaller diurnal cycle and often shows characteristics similar to convection over tropical oceans, with little lightning activity and less intense cloud cells. In contrast, convection during the break periods is often isolated
13
and localized predominately over land, generally initiated by sea breezes and other local circulations, but associated cloud systems are more intense, taller and electrically more active with a stronger diurnal cycle in the afternoon. However, using both temperature and wind profiles from radiosondes, Pope et al. (2009a) showed that at least five distinct atmospheric regimes are needed to explain the main modes of atmospheric variability over Darwin during the wet season (September–April). These local regimes differ significantly in their synoptic environments, cloud pattern, rainfall distributions, ice-cloud properties, convective cloud properties and their diurnal evolution (Pope et al. 2009a; Protat et al. 2011; Kumar et al. 2013). Similarly, northern Australia also exhibits a high intraseasonal variability during the Australian summer monsoon (Drosdowsky 1996; Wheeler and McBride 2011). The Madden–Julian oscillation (MJO) is one of the major sources of intraseasonal variability and has a local intraseasonal period of 30–90 days (Zhang 2005), providing the strongest modulation of synoptic conditions over northern Australia in austral summer (Wheeler et al. 2009; hereafter WH09). Wheeler and Hendon (2004; hereafter WH04) showed a more than three times increase in the probability of extreme weekly rainfall in northern Australia during the convectively active MJO phase compared to the suppressed phase. Pope et al. (2009b) found a substantial increase in the total number of mesoscale convective systems (MCSs) during the active conditions of MJO. However, how the rainfall over different surface types (land, islands and oceans) in the vicinity of Darwin is modulated on various temporal scales by the MJO phases has not been investigated yet. Hence, the main objective of this work is to characterize the temporal and spatial variability of rainfall over the Darwin and its vicinity using datasets of unprecedented spatial and temporal resolutions, with particular emphasis on the relationship to phases of the MJO. These statistics may be used as a testbed to explore many aspects of the model performance. Furthermore, these results may also be used as a guide to plan future field experiments around this region. One such field experiment where these statistics have been useful is the High Altitude Ice Crystals–High Ice Water Content (HAIC–HIWC) Project (Protat et al. 2014). In addition to the above objectives, this study also intends to investigate the physical mechanism behind the occurrence of bimodal peaks in the diurnal cycle of rainfall over Darwin as observed in the study of Rauniyar (2006) and shown in Fig. 2. The study used one DJF (December– February) season data from the Enhanced Observing Period 3 (EOP3; 01-Oct-2002 to 30-Sept-2003) of the Coordinated Enhanced Observation Period (CEOP; Koike 2004). A semidiurnal cycle in the frequency of area-averaged rainfall over land surrounding Darwin has also been reported previously (e.g., Soman et al. 1995). Furthermore, the bimodal
Spatial and temporal variations in rainfall over Darwin and its vicinity during different… 2.50
Observation vs Models at Darwin
Normalized rainrate
2.25 2.00 1.75
●
1.50 1.25
In situ BMRC_F BMRC_A UKMO JMA NCEP
● ●
●
● ●
●
●
1.00
●
●
●
●
● ●
● ●
●
0.75
● ●
0.50
●
●
●
● ● ●
0.25 00
03
06
09
12
15
18
21
●
24
Local statndard time (hr)
Fig. 2 Comparison of hourly mean diurnal cycle of observed rainfall with the model simulated diurnal cycle at Darwin, Australia using a single DJF (2002/03) season data. The diurnal cycles are normalized by their respective means. From Rauniyar (2006)
feature has also been observed at several other tropical coastal sites during rainy seasons (Rauniyar et al. 2007; Santos e Silva et al. 2009; Kanamori et al. 2013). However, the physical mechanism behind such occurrence is not known. The diurnally regulated surface solar radiative heating is a prime mechanism behind the occurrence of mid to late afternoon rainfall over land; however there are various explanations for early morning rainfall maxima over land (Yang and Smith 2006). The present study intends to determine under which circumstances Darwin experiences more rainfall during the early morning (0200–0400 LST). The paper is organized as follows: a brief description of data, different indices used to classify the large-scale circulations and methodology is given in Sect. 2. The variations in the daily rainfall climatology and associated dynamics and thermo-dynamics as a function of the large-scale circulations are shown in Sect. 3. The mechanism behind the occurrence of an early morning rainfall maximum over Darwin is described in Sect. 4. The characteristics of the diurnal cycle of rainfall over different surface types during different large-scale conditions are described in Sect. 5, while Sect. 6 summarizes the major findings of this study, discusses possible explanations for the observed variations and suggests future work.
over land adjacent to Darwin, as shown in Fig. 1. The stations shown as filled square in Fig. 1, are operated by the Australian Bureau of Meteorology (BoM) and provide daily accumulated rainfall (mm day−1). However, only 7 stations have less than 10 % missing values during the 16 DJF (1995/96–2009/10) seasons and hence are used in this study. Furthermore, only two stations have hourly data with an uncertainty of ±0.1 mm h−1: a tipping-bucket rain gauge located at Darwin Airport (12.4239°S, 130.8925°E; white filled square in Fig. 1) and operated by the BoM and an optical rain gauge located (12.425°S, 130.892°E; white filled circle in Fig. 1) adjacent to the BoM site and operated by Atmospheric Radiation Measurement (ARM) Climate Research Facility (ACRF). At the time of study, the hourly resolution rainfall data were available only for the 7 DJF (2001/02–2007/8) seasons from the BoM site and only for the 8 DJF (2002/03–2009/10) seasons from the ACRF site. Less than 2 (1) % of data are missing at BOM (ACRF) site, at which only 3 (1) days have more than 50 % of missing hourly values. The second data source is a 5.3 cm wavelength C-band polarimetric (CPOL) radar located at Gunn Point (12.25°S, 131.04°E, filled triangle in Fig. 1), approximately 20 km northeast of Darwin (Keenan et al. 1998). The CPOL radar performs a volumetric scan every 10 min within a 150 km scan radius covering land, coastal, islands and ocean regions (black circle in Fig. 1). The rain rate is estimated using a multiparameter rainfall algorithm described in Bringi et al. (2001, 2004) and is available in a three dimensional Cartesian grid of 2.5 km resolution horizontally and 0.5 km resolution vertically, known as Constant Altitude Plan Position Indicators (CAPPI). In this study, hourly accumulated rain rate at 2.5 km CAPPI and in a 20–120 km radial range is used to avoid errors in estimation due to limited sample size at close ranges caused by radar’s inability to scan directly overhead and at large ranges due to beam spreading (May et al. 2012). Only 8 DJF (1998/99– 2006/07; excluding 2000/01) seasons of CPOL rainfall data were available at the time of study. 2.2 Other datasets
2 Data and methods The study has used multiple years of high-resolution spatial and temporal data from various sources as discussed below to minimize the effect of single large events on overall statistics. 2.1 Observed and model rainfall The present study utilizes various sources of rainfall data. The first source is from several rain gauges all located
The study has used the daily 2300 UTC (0830 LT) radiosonde observations to examine the atmospheric and thermodynamic conditions over Darwin during different synoptic conditions. The 2300 UTC data are used to avoid the influence of strong afternoon convection on wind and moisture profile. Similarly, the hourly average surface air temperature at 2 m and wind at 10 m are also used to compute the daily maximum temperature and to determine existence of the sea breezes. These data were obtained from the ARM Climate Modeling Best Estimate (CMBE) atmospheric and radiation datasets (Xie et al. 2010) and were available
13
S. P. Rauniyar, K. J. E. Walsh 20 10 −1 ms
Clim DE E DW SW ME
ME SWDW DE E
ME
Height above ground (km)
15 E
ME DE SW DW
E
ME DE SW
10 E
DW ME
EDE
SW
5
ME
DW
E
ME DE
SW
DW
E
ME
(b)
(a) 0 DE
E
DW
SW
ME
Darwin regimes
20
30
40
50
60
70
80
90
Relative humidity (%)
Fig. 3 The profiles of a daily mean horizontal winds (ms−1) and b daily mean relative humidity (%) as a function of the Darwin atmospheric regimes at 0.5 km vertical resolution using the radiosonde measurement at Darwin, Australia. The radiosondes were obtained from the ARM Climate Modeling Best Estimate (CMBE) atmospheric and radiation datasets (Xie et al. 2010) and were available for 8 DJF (2002/03–2009/10) seasons. The scale of the vectors is printed on top right corners of left panel and the north direction points
upward. The dark black arrows indicate wind vectors that are statistically significant at the 95 % level compared to climatology according to the t test either in the zonal or in the meridional component. The climatological profile of relative humidity is shown shaded and the thick relative humidity lines indicate the layer where the humidity profiles are statistically significant at the 95 % level according to the t test
only for 8 DJF (2002/03–2009/2010) seasons at the time of study. In addition, the daily optimum interpolated sea surface temperature (SST) at 0.25° spatial resolution (Reynolds et al. 2007) has also been used for the same period to better understand the impact of propagating MJO on local land–sea thermal contrast near Darwin.
profiles of these regimes using CMBE radiosonde observations are shown in Fig. 3. Both the DE and E regimes exhibit south-easterly winds in the low- and mid-troposphere and have a very dry moisture profile throughout the troposphere. As these two regimes mostly occur during the pre-monsoon and the post-monsoon seasons (Pope et al. 2009a), their occurrence are very low, respectively 1 and 7 % during DJF seasons (Fig. 4a). In contrast, the DW regime has the largest values of relative humidity throughout the atmosphere with a westerly zonal wind up to about 7 km height and easterly above this level. This regime corresponds to the active monsoon periods and satisfies the wind-based definition of monsoon onset of Drosdowsky (1996). The SW regime exhibits a shallow westerly wind below about 2 km height, with weak easterly winds above that level and has a slightly drier atmosphere than the climatology. In the ME regime, the lower tropospheric zonal winds are easterly, but are weaker than in the DE or E regimes. However, the zonal easterly extends throughout the entire troposphere and has a slightly wetter atmosphere than the climatology. This regime is associated with break periods in the monsoon (May et al. 2008). The DW, SW and ME regimes are more common during DJF seasons
2.3 Indices for large‑scale circulations In general, two different approaches can be used to study the rainfall variability over Darwin: the Darwin atmospheric regimes as identified by Pope et al. (2009a) and the MJO phases as proposed by WH04. These are explained in brief as follows: 2.3.1 Darwin atmospheric regimes To better explain the variability of the north Australian wet season, using a long-term record of radiosonde observations over Darwin, Pope et al. (2009a) objectively derived five distinct atmospheric regimes: Dry Easterly (DE), Easterly (E), Deep Westerly (DW), Shallow Westerly (SW) and Moist Easterly (ME). The wind and relative humidity
13
Spatial and temporal variations in rainfall over Darwin and its vicinity during different… 0.45 0.40
(a)
●
Probability
0.35 0.30 0.25
●
●
0.20 0.15 0.10
●
●
0.05 0.00
●
DE
DW
E
La Nina El Nino Clim.
SW
ME
Darwin regimes 0.40
Probability
0.35
(b) ●
0.30 0.25 0.20 0.15 0.10 0.05
●
● ●
2.3.3 Inter‑annual variation and inter‑relationship
●
●
●
●
●
0.00 1
2
3
4
5
6
7
8
W
MJO phases 0.80
Probability
0.70
DE DW E SW ME
(c)
0.60 0.50 0.40 0.30 0.20 0.10 0.00 1
2
3
MJO (RMM) index developed by WH04 without applying any temporal filter has been used to discriminate the MJO phases. The index is based on the leading pair of combined empirical orthogonal functions (EOF) analysis of nearequatorially (15°S–15°N) averaged zonal winds at 850 and 200 hPa and outgoing longwave radiation (OLR). When the MJO is in a strong cycle (amplitude greater than 1), deep convection is shown to propagate eastward from the Indian Ocean (phases 2–3) to the MC (phases 4–5) and over the western Pacific (phases 6–7), before decaying around the dateline in the central Pacific (phases 8 and 1; WH04). Figure 4b shows that the frequencies of the MJO phases during the DJF season range from 5 to 10 %, except for the weak MJO phase, which occurs about 35 % of the time. The weak phase climatology is not shown in this paper as it showed features very similar to the climatology.
4
5
6
7
8
W
MJO phases
Fig. 4 The climatological probability of occurrence of a each regime and b each MJO phase using 36 DJF (19974/75–2009/10) seasons (short-dashed line). The same, but for the DJF seasons classified as La Niña (12 DJFs; filled circle on gray solid line) and El Niño (11 DJFs; open circle on black solid line) phases of El Niño Southern Oscillation (ENSO) based on the Niño 3.4 index. c Probability of occurrence of each regime in each of the MJO phases
with frequency of occurrence being respectively 20, 25 and 40 %, (Fig. 4a). 2.3.2 Real‑time multivariate MJO (RMM) index As the MJO involves planetary-scale regions of enhanced and suppressed convection/rainfall (Madden and Julian 1994), the intraseasonal signals of the MJO can be extracted by applying a time bandpass filter. However, in this study, a seasonally independent Real-time Multivariate
Since, the rainfall over Darwin experiences a large interannual variation, it is necessary to diagnose whether the above two indices (Darwin regimes and the MJO phases) also experience the same variability. This has been achieved by calculating the frequencies of these indices separately for the El Niño and La Niña phases of El Niño Southern Oscillation (ENSO) using the Niño 3.4 index (www.cpc. ncep.noaa.gov) from 1974/75 to 2009/10 DJF seasons. The result shows that the Darwin regimes exhibit a large interannual variation (Fig. 4a). The frequency of DW regime decreases substantially during El Niño compared to La Niña. This is compensated by an increase in the frequencies of the SW and E regimes during El Niño. The reverse is observed during La Niña, which is expected as the rising branch of the Walker circulation is mainly located over the Maritime Continent during La Niña. In contrast, as the inter-annual variability associated with ENSO is effectively removed before calculating the RMM indices (WH04), no significant inter-annual variation exists in the MJO phases, except in phase 5. Phase 5 shows a slight decrease in frequency during El Niño compared to La Niña (Fig. 4b). This may be due to a higher occurrence of the DW regime in phase 5, whose frequency substantially decreases during El Niño condition (Fig. 4a). In contrast, the MJO phases 1–3 experience a higher contribution from the ME regime compared to other regimes (Fig. 4c) as the active center of convection is located east of the study region during these phases. The rest of the MJO phases (4–8) are found to be a mixture of the DW, SW and ME regimes. As expected, the contribution from the pre-monsoonal DE and E regimes is very low in all the monsoonal MJO phases shown here. These results show that the rainfall climatology based on Darwin regimes can be substantially modulated by ENSO conditions and hence the variability of rainfall cannot be
13
S. P. Rauniyar, K. J. E. Walsh
studied by using only a few seasons of data. In contrast, the MJO phases may be a better discriminator of large-scale circulations as they are derived using large-scale winds and OLR. One could develop the rainfall climatology based on the MJO phases with fewer seasons of data because as Fig. 4b shows inter-annual variability of MJO phase is low. Furthermore, the propagating nature of the MJO enables the construction of the life-cycles of rainfall-building mechanisms (see Sect. 3.4) which is not possible by using the Darwin regimes as unlike the MJO phase definition they do not show any form of propagation or quasi-periodic pattern of occurrence.
and the Darwin regimes. Furthermore, to explore the variations in diurnal cycle of rainfall over different surface types as a function of large-scale circulations, the hourly mean diurnal cycles are composited by simple averaging of CPOL rainfall. An equally weighted 3-point running mean is applied to reduce noise in the hourly mean diurnal cycle due to under-sampling. Similarly, a 5-point boxcar filter in X- and Y-directions is used to smooth the spatial inhomogeneity. Harmonic analysis is also applied on the hourly mean diurnal cycle of CPOL rainfall to identify the regions of dominant semi-diurnal variability.
2.4 Methodology
3 Daily climatology as function of the MJO phases
The composites of rainfall and other atmospheric and thermodynamic variables are computed for each regime and also for each MJO phase, including the weak phase. Statistical significance for the composite daily rainfall anomalies and event probabilities of the MJO phases is evaluated using a non-parametric resampling approach as discussed in WH09. The advantage of this method is that it makes no assumptions about the normality of the data. In this method, the MJO index is successively shifted by a regular interval (6 days) relative to the time series of rainfall for an arbitrary 400 times. For each shift, the rainfall anomalies and relative frequencies are recomputed for each MJO phase. This generates a null distribution and its 5 (2.5)th and 95 (97.5)th percentiles values are used as the thresholds for significance at the 10 (5) % level for a two-tailed test. The other variables (e.g., wind, temperature, moisture) are generally normally distributed and hence the Student’s t test (hereafter t test) is applied to judge the statistical significance at 95 % level compared to climatology. The hourly diurnal cycle in local time is computed for each MJO phase and also for each regime. However, the composite diurnal cycles as a function of the MJO phases showed larger noise due to the limited number of rain events at each hour. Hence, EOF analysis is applied on the hourly in situ observation in an innovative way to identify the most dominant diurnal patterns and their associated large-scale circulations. This was achieved by arranging the in situ observations in a matrix form such that a row represents the hourly observations of a particular day (total number of rows = n days) and a column represents the time-series of rainfall at any particular hour (total number of columns = 24). The anomaly data matrix is created by subtracting the mean diurnal cycle from each row which is then decomposed using singular value decomposition (SVD). The non-degeneracy of the EOF patterns is evaluated using a rule of thumb (North et al. 1982). Finally, the time-series of significant principal components (PC) are binned separately into the corresponding MJO phases
The characteristics of Darwin regimes for various variables (e.g., wind, moisture, rainfall, cloud, etc.) over Darwin and its surrounding regions have already been shown in previous studies (e.g., Pope et al. 2009a; Protat et al. 2011; Kumar et al. 2013). Our results are also found to be very similar to these previous studies and hence are not shown here. Hence, only the results composited according to the MJO phases are shown in this section.
13
3.1 Wind and humidity profile To determine the dynamical and thermo-dynamical features during the MJO phases, the vertical profiles of horizontal winds and relative humidity as function of the MJO phases are computed and are shown in Fig. 5. Phase 1 exhibits zonal easterly winds of varying magnitudes throughout the troposphere (Fig. 5a). In phase 2, the magnitude of low- to mid-level zonal easterly winds are slightly strengthened, however these winds then turn into weak westerly winds between 7 and 12 km height and above which the easterly jet prevails. From phase 3, the magnitude of low- to midlevel zonal easterly winds start to decrease. These winds change to weak northwesterlies between 5 and 10 km height and above 10 km these winds turn into southeasterlies and easterlies. The moisture profiles during these phases show slightly drier atmosphere than climatology (Fig. 5b). From phase 4, the reversal of wind takes place and the low- to mid-level easterly winds are completely replaced by weak westerly winds (Fig. 5a) with a substantial build-up of moisture in the atmosphere due to advection of warm moist air from the adjacent seas (Fig. 5c). Hence, the phenomena known as “preconditioning” takes place in phase 4. In phase 5, the magnitudes of westerly winds further strengthen creating the wettest atmosphere of all phases and herald the actual arrival of the convectively active MJO over Darwin. In phase 6, although the westerly winds become stronger, the humidity starts to decrease making the atmosphere only slightly wetter than the
Spatial and temporal variations in rainfall over Darwin and its vicinity during different… 20
Height above ground (km)
10 −1 ms
15
10
5
(a)
0
1
2
3
4
5
6
7
8
Clim.
MJO phase 20 Ph_1 Ph_2 Ph_3 Ph_8
13
Height above ground (km)
32 81 15
67 5 4 75 46 7 5
3 1 82 13
7 46 5
8 123
10
Ph_4 Ph_5 Ph_6 Ph_7
75
1 3 82 13
7
64 7
13 8 2 1 3
(b)
0 20
30
40
50
60
5
7 6 54
3 8 12
5
5
70
Relative humidity (%)
80
7 5 64 5 7
(c) 20
30
40
50
60
70
80
upper-level zonal easterly winds first increase and then decrease; however the strongest upper-tropospheric easterly winds exist in phase 5. Phase 8 exhibits southeasterly winds of continental origin from 1 km up to 8 km height, resulting in the driest atmosphere of all phases (Fig. 5b). We also analyzed the hourly average surface winds at 10 m observed at Darwin by constructing the wind rose plots according to the MJO phases to identify the sea breezes dominated MJO phases (Fig. 6). The wind rose plots show that Darwin experiences westerly and northwesterly winds more than 50 % of the time in the afternoon (1230–1530 LT) during all MJO phases. However, the morning (0330–0630 LT) wind rose plots show that easterly and southeasterly winds are stronger only during the MJO phases 8 and 1–3. The previous studies (e.g., Crosman and Horel 2010; Miller et al. 2003; Porson et al. 2007; Arritt 1993) have shown that the sea breezes are suppressed when the large-scale flow has the same direction, unless the large-scale flow is very weak. Similarly, the sea breezes are also suppressed when the opposing flow becomes very strong. In contrast, the most intense sea breezes develop with inland penetration for calm to moderate opposing synoptic flow, with no inland sea breeze penetration for strong opposing large-scale flow. As the large-scale flow and afternoon surface winds blow from the same direction during the MJO phases 4–7, one can conclude that the suppression of sea breezes over Darwin is highly likely during these phases. This is also confirmed by the existence of weaker land breezes in the morning (0330–0630 LT) during these phases. In contrast, the large-scale flow and afternoon surface winds blow from opposing direction during the MJO phases 8, 1–3 which enhances sea breezes during afternoon as a result strong flow reversal occurs during the morning as land breezes. However, the magnitude of localscale land–sea thermal contrast also plays an important role in the formation or suppression of land–sea breezes, as explained below.
Relative humidity (%)
3.2 Land–sea thermal contrast Fig. 5 The same as Fig. 3, but according to the MJO phases. b The relative humidity profiles for the MJO phases 8 and 1–3 whereas c shows the same for the MJO phases 4–7
climatology. Although phase 6 is more dominated by the DW regime (Fig. 4c), both the MJO phases 5 and 6 satisfy the wind-based definition of monsoon onset of Drosdowsky (1996). However, their moisture profiles are quite different (Fig. 5c) suggesting that a monsoon onset definition based on wind only is not sufficient. In phase 7, the westerly winds are the strongest in magnitude, but are limited to a shallow depth of 2 km only. Furthermore, during this phase, the atmosphere remains slightly drier than average. In addition, during phases 4–7, the magnitude of mid- to
This section explores the variation in the local-scale thermal contrast between Darwin and its surrounding sea by the passage of MJO. The average value of SST inside the radar domain consistently remains above 30 °C during the DJF season. The warmest SST is generally distributed along the coast of the mainland and over the Van Diemen Gulf (lower right panel of Fig. 7). As the sea breezes initiate in the afternoon when the land is comparatively warmer than its adjacent sea, we computed the likelihood of sea breeze formation over Darwin during the MJO phases as follows. The daily maximum surface air temperature observed at Darwin is first subtracted from the daily mean SST at each grid box which is then averaged according to
13
S. P. Rauniyar, K. J. E. Walsh 8)
W
N
●
1)
E
W
N
●
2)
E
W
N
●
3)
E
W
N
●
0−2
E
2−4 4−6
S 4)
W
S
N
●
5)
E
W
S
N
●
6−14
6)
E
W
N
●
S 7)
E
W
N
●
0−2
E
2−4 4−6
S 8)
W
S
N
●
1)
E
W
S
N
●
6−14
2)
E
W
N
●
S 3)
E
W
N
●
0−2
E
2−4 4−6
S 4)
W
S
N
●
5)
E
W
S
N
●
6−14
6)
E
W
N
●
S 7)
E
W
N
●
0−2
E
2−4 4−6
S
S
S
6−14
S
Fig. 6 Wind rose plots according to the MJO phases showing predominant directions and speeds of hourly average surface winds at 10 m in the afternoon (1230–1530 LT; first two rows) and in the morning (0330–0630 LT; last two rows) using 8 DJF (2002/03–
2009/10) seasons of the ARM–CMBE data. The concentric circles represent wind frequency at an interval of 10 % with outer circle at 50 %
the MJO phases, assuming no significant diurnal variation in SST. This way the negative (positive) values over the seas signify that the land near Darwin is comparatively warmer (cooler) with a high probability of sea breezes initiation (suppression) during afternoon. During phases 4–6, a minimum land–sea thermal contrast exists as the surrounding daily average SSTs are almost the same as surface air daily maximum temperature over Darwin. Hence, during these phases, the sea breezes will be suppressed even without existence of any strong large-scale flow. In fact, during phase 5, the SST over the Van Diemen Gulf is warmer than adjacent land even during the afternoon. In contrast, during the remaining phases (1–3 and 7–8), the afternoon surface temperature over land is found to be higher than adjacent seas, thus initiating the sea breezes. This is clearly related to variations in incoming solar radiation at the surface: Protat et al. (2011) found a higher occurrence of optically thick clouds during phases 5 and 6 which causes a significant
decrease in the net short wave radiation at the surface. May et al. (2012) showed that the surface receives only about 40 % of the clear-sky amount of radiation during the active monsoon conditions (DW regime or MJO phases 5 and 6). In contrast, they found that the surface receives up to 80 % of the clear-sky amount due to the modest amounts of cloud cover during the break conditions (ME regime or MJO phases 1–3 and 8). Hence, the above analysis of dynamic, thermo-dynamic profiles, surface winds and the local-scale land–sea thermal contrast showed that the seabreezes are suppressed during the MJO phases 4–7 whereas they are enhanced during the phases 1–3 and 8.
13
3.3 Rainfall and its spatial variation The mean daily rainfalls over the period of analysis over inland areas of Darwin at several of the in situ sites (Fig. 1) are observed to be around 13 mm day−1. These values are
Spatial and temporal variations in rainfall over Darwin and its vicinity during different…
Fig. 7 The first 8 maps show the composites of maximum land–sea thermal contrast according to the MJO phases 1–8 calculated by subtracting the diurnal (daily) maximum near surface air temperature recorded at Darwin from the daily mean SST. The warm (cool) colors represent the warmer (colder) SST regions compared to Dar-
win and hence suppressed (enhanced) conditions for development of sea breezes. The average of 8 DJF (2002/03–2009/10) seasons SST is shown at the lower right corner. The number of days in each MJO phase and climatology are shown in the parentheses at the lower right corner of each map
consistent with the longterm average rainfall of the northwest top end (Darwin) of Australia (www.bom.gov.au). However, these values oscillate heavily during the passage of the MJO as shown in Fig. 8a. In general, each strong MJO phase (event) lasts 4–6 days and a typical life-cycle of MJO is about 50 days (WH09). The results shown in Fig. 8 are based on average of 18–26 unique MJO events, except for the weak phase (46 events). These statistics are based on 16 DJF (1995/96–2009/10) seasons of data. Both the coastal (within 50 km from the coast; C in Fig. 8a) and inland (more than 50 km away from the coast; L in Fig. 8a) sites show very similar variations in rainfall during the MJO phases. In general, during phases 4–6, above normal rainfall occurs whereas it is below normal in the remaining strong MJO phases. The rainfall anomalies are however statistically significant at the 95 % level only for
phases 5 and 2, when the highest and the lowest rainfall are observed, respectively. Between these two phases, the average difference is more than 10 mm day−1 which is about 75 % of the mean. This value is roughly the same as reported by Hendon and Liebmann (1990b) who found that the amplitude of rainfall anomaly oscillates by more (less) than 4 mm day−1 during the enhanced (suppressed) phase of MJO. Since the daily rainfall anomaly is very sensitive to an outlier event compared to the probability of rainfall exceeding a certain threshold (WH09), we computed the event probabilities as well to verify whether the other metrics show similar oscillations or not. We have defined the event probability as the probability of rainfall exceeding the climatological median rainfall; however any meaningful value can be used as a threshold. On average,
13
S. P. Rauniyar, K. J. E. Walsh
Rain anomaly (mm day−1)
10 8
014015_C 014227_C 014192_L 014219_L 014222_L 014226_L 014183_L
(a)
6 4 2 0 −2 −4 −6 −8 −10
Event probability (%)
80
13.7 (3.6) 13.6 (3.0) 13.6 (4.4) 14.1 (4.6) 12.8 (3.8) 13.5 (4.4) 12.5 (3.9)
(b)
70 60 50 40 30 20 1
2
3
4
5
6
7
8
W
MJO phases
Fig. 8 The composites of a daily mean rainfall anomalies and b the probabilities of daily rainfall greater than the median rainfall for each MJO phase using 16 DJF (1995/96–2009/10) seasons data from seven BoM rain gauges stations as shown in Fig. 1. The upward and down‑ ward filled triangles indicate the signals that are statistically significant at the 95 % level. The letters C and L in the legends represent respectively, coastal (with in 50 km from the coast) and land sites. The average of daily mean rainfall using all days is shown in the lower panel with the corresponding median in the parentheses in the legend
the probability of exceeding the median rainfall value is above 60 % during phases 4–6, while it is below 40 % during phases 2 and 8. The event probability shows a very similar variation as observed for the rainfall anomaly, except for phase 6 (Fig. 8b). In phase 6, the event probability is the highest and statistically significant at the 95 % level whereas the highest positive rainfall anomaly occurs in phase 5. Further analysis shows that the rainfall events are more frequent but are less intense in phase 6 compared to the phase 5 (not shown). Hence, both metrics show that the area near Darwin experiences an increase in rainfall during the enhanced phase of the large-scale circulation (phases 4–6) while a general decrease occurs in rainfall during the suppressed phase of the large-scale circulation (1–3 and 8). Furthermore, to determine whether the same or different impact of MJO is felt over land, coastal, islands and ocean regions in the vicinity of Darwin, the daily rainfall
13
anomalies and mean at each valid pixel of CPOL are computed as function of the MJO using the 8 DJF (1998/99– 2007/08, except 2000/01) seasons data (Figs. 9, 10). The daily climatology of CPOL estimated rainfall shows a value just above 8 mm day−1 at Darwin (lower right panel of Fig. 9). The value is slightly lower than the in situ average of the same period (11 mm day−1). This is a known problem with ground-based radar observations like CPOL due to the radar scanning pattern and variability of the drop size distributions (Grimsdell et al. 2010). As each CPOL scan takes 10 min to complete, it is quite possible that the CPOL may miss short-term high intensity rainfall events. In addition, the spatial smoothing applied here also contributes partially in this underestimation. Nevertheless, the CPOL rainfall anomalies over Darwin according to the MJO phases show similar variations to those seen in Fig. 8a, even though the CPOL composites are based on about 10–12 unique MJO events, half of the events used for the in situ stations as discussed in the above paragraph. Furthermore, the peak to trough difference between active and suppressed phases (>6 mm day−1) in this case is also above 70 % of the mean, which is close to the in situ value shown in the previous paragraph. In addition, the rainfall anomaly maps also show a large and distinct spatial variation in rainfall during the MJO phases (Fig. 9). During phases 5–6, above normal rainfall occurs over most parts of the radar domain whereas below normal rainfall occurs during the remaining phases. However, there exist a few notable exceptions to this general rule. For example, over the Tiwi Islands, the highest positive rainfall anomaly exists in phase 2 which is opposite of what is observed over other parts of the radar domain where negative rainfall anomalies dominate. The Tiwi Islands receive most of their daily rainfall from deep convective storms, which form almost every afternoon and are locally known as “Hectors” (e.g., Keenan et al. 1989, 1990; Skinner and Tapper 1994; Beringer et al. 2001; Crook 2001; Oliphant et al. 2001). The Hectors are found to be most active during the transition season or during the build and break periods of the Australian monsoon. However, none of the previous studies have shown the relationship between the Hectors and the MJO phases. The probable reasons behind the occurrence of intense Hectors in phase 2 are discussed in Sect. 5. The other notable within-phase spatial anomaly exists in phases 6 and 4. During phase 6, above and below normal rainfall exists over the southwestern and southeastern corner of the radar domain, respectively. In contrast, during phase 4, the pattern has almost opposite sign, but has a weaker magnitude. These distinct spatial patterns are related with the diurnal evolution of rainfall, which is influenced by the westward propagating squall lines from Cape York Peninsula and is discussed in more detail in Sect. 5.
Spatial and temporal variations in rainfall over Darwin and its vicinity during different…
Fig. 9 The first 8 maps show daily mean rainfall anomalies (mm day−1) from CPOL for the MJO phases 1–8. The hatched regions are statistically significant at the 90 % confidence level according to bootstrap test. The rainfall climatology in mm day−1
using 8 DJF (1998/99–2006/07; excluding 2000/01) seasons is shown at the lower right corner. The number of days used in the analysis is shown in the parentheses at the lower right corner of each map
3.4 Life cycles of rainfall building mechanisms
a general increase in rainfall over other parts of the radar domain. The large-scale mechanism peaks in phase 5 producing a widespread above normal rainfall over the entire study domain due to an actual arrival of the convectively active MJO over Darwin. However, even during this phase there exists a clear west-east gradient in rainfall, with most rainfall occurring over the western side (i.e., Beagle Gulf) compared to the eastern side (i.e., Van Diemen Gulf) of the radar domain. May et al. (2012) have found a similar spatial distribution of rainfall during the active monsoon. They hypothesized that the strong modification of the boundary layer by the upstream islands is responsible for comparatively less rainfall on their eastern side. From phase 6, the
The above findings suggest that there are competing influences between two different mechanisms: the large-scale forcing and the mesoscale sea-breeze mechanism over the study domain. A complete life-cycle of these two mechanisms and their association with the MJO phases can be hence established due to the propagating nature of the MJO (Fig. 10). The large-scale forcing starts to gradually build up from phase 4 when the reversal of easterly winds into westerly winds takes place at low-level. This moistens the atmosphere, but suppresses the sea breezes resulting into a large decrease in rainfall over the Tiwi Islands and
13
Avg. Rainrate (mm hr−1)
S. P. Rauniyar, K. J. E. Walsh 1.0
(a)
BoM site
0.8 ●
0.6
● ●
● ●
0.4
●
●
●
● ●
● ● ●
●
● ●
●
●
● ● ●
● ●
●
●
0.2
Avg. Rainrate (mm hr−1)
0.0 1.0
(b)
0.8 0.6
●
● ● ● ●
ACRF site ●
●
● ● ●
●
● ●
●
0.4
●
● ●
●
●
● ●
● ●
●
●
0.2
Avg. Rainrate (mm hr−1)
0.0 1.0
(c)
CPOL
0.8 ●
0.6
●
0.4
● ● ●
● ●
● ● ●
● ●
0.2
●
●
●
● ● ●
●
●
● ● ●
●
21
24
●
0.0 00
03
06
09
12
15
18
Local time (hr)
Fig. 11 The hourly mean diurnal cycle of total days rainfall a at the BoM site using 7DJF (2001/02–2007/08) seasons, b at the ACRF site using 8 DJF (2002/03–2009/10) seasons and c at the CPOL grid close to Darwin using 8 DJF (1998/99–2007/08; excluding 2000/2001) seasons of data. The error bar shows the 95 % confidence level of hourly mean rainfall based on bootstrap test
Fig. 10 The left panel shows the composites of daily mean rainfall (mm day−1) from CPOL during enhanced large-scale conditions (e.g., MJO phases 4–7) while the right panel shows the same, but during enhanced sea breezes conditions (MJO phases 8, 1–3)
large-scale rainfall starts to decline due to the eastward shift of the main convection center associated with the MJO. However, this decrease starts to affect the eastern part first as the maximum is found to be very much still located on the western side. The exact mechanism behind this may be explained using high resolution numerical modeling which is beyond the scope of this paper. In phase 7, the rainfall further decreases, but still has the characteristics of largescale rainfall as no amplification of sea breezes is observed over the Tiwi Islands or over the mainland. From phase
13
8, the reinvigoration of sea breezes takes place and as a result the rainfall is predominately concentrated over the islands and the mainland only. The strength of sea-breezes becomes larger in phase 1, and reaches its maximum during phases 2 and 3 as seen in a higher occurrence of stronger land breezes over Darwin resulting in more isolated rainfall over land. Over the Tiwi Islands, this has been observed in a significant increase in the frequency of Hector development producing above normal rainfall in phase 2. However, in phase 3 over the Tiwi Islands, the strength of sea breezes starts to weaken as a result below normal rainfall is observed. These results show that the strong MJO phases over Darwin can be broadly classified into two categories: the active and suppressed MJO which is consistent with the findings of WH09. The active MJO consists of the phases 4–7 when the rainfall is produced by the large-scale forcing
Spatial and temporal variations in rainfall over Darwin and its vicinity during different…
and covers both land and ocean regions. In contrast, the suppressed MJO consists of the phases 1–3 and 8 when the rainfall is generated by the local-scale land–sea breezes and predominately located over the land/islands only. However, there exists a substantial regional variation within the phases that are classified into an active or suppressed MJO as shown in this study.
4 Mechanism behind the early morning rainfall maxima Figure 11 shows the mean diurnal cycle of total rainfall near Darwin from three different sources of rainfall observations. Even though the composites are based on different time periods, all diurnal cycles show the existence of two dominant peaks, These primary and secondary peaks occur during late afternoon/early evening (1500–1700 LT) and during early morning (0200–0400 LT), respectively. In this section, we have applied EOF analysis in an innovative way to determine the large-scale conditions responsible for the occurrence of the secondary diurnal peak at Darwin. As discussed in Sect. 2, only two in situ sites (BoM and ACRF) have hourly observations of rainfall near Darwin. We applied EOF analysis to these two sites separately and found that they produce very similar results. Hence, for brevity, we have only shown the results of EOF analysis applied on the average of hourly observations computed using the common period of 6 DJF (2002/03–2007/08) seasons in Fig. 12. The results show that only the first two leading EOF modes are non-degenerate from the rest, accounting for more than 27 % of the total variance (Fig. 12a). EOF1 explains more than 15 % of the total variance and shows a diurnal cycle with multiple peaks (Fig. 12b). However, the strongest peak exists between 0200–0400 LT. EOF2 explains about 12 % of the total variance and shows a strong diurnal cycle with an afternoon peak at 1600 LT. The composites of the corresponding PCs when binned according to the large-scale circulations show that the amplitude of PC1 is stronger during MJO phases 4–6 (Fig. 12e) or during the DW regime (Fig. 12f) whereas the amplitude of PC2 is larger during MJO phases 1–3 and 8 or during the ME regime. The mean diurnal cycles composited using the rainfall from the MJO active (i.e., phases 4–6) days and MJO suppressed (i.e., phases 1–3 and 8) days show similar diurnal cycles as in EOF1 and EOF2, respectively (not shown). However, compared with simple averaging, the EOF approach has the advantage that it does not require a prior knowledge about the state of large-scale conditions; rather it detects the common diurnal patterns and their associated large-scale conditions. Furthermore, to explore whether the bimodal feature is tied to Darwin or has any spatial preference, Fourier
decomposition (harmonic analysis) is performed on the mean diurnal cycles of CPOL rainfall composited according to the MJO phases. The first harmonic represents the diurnal cycle with one peak whereas the second harmonic represents the semi-diurnal cycle with two peaks at 12 h apart. Only the result from phase 5 is shown in Fig. 13 as it has the strongest relationship with the early morning peak (Fig. 12e). Note that the regions with amplitude less than 0.1 mm h−1 in Fig. 13 are whitened. The result shows that the semi-diurnal cycle dominates along the southwest coast near Darwin where it accounts for more than 50 % of total variance. In contrast, over the same region, the first harmonic (diurnal cycle) explains less than 20 % of total variance. The spatial distribution of phase of second harmonic (time of maximum rainfall) shows that the rainfall over the southwest coastal region near Darwin occurs during early morning (0200–0400 LT) due to an eastward advection or expansion of rainfall systems that form over the Beagle Gulf during mid-night (0000–0200 LT).
5 Spatial variations in the diurnal cycle of rainfall This section explores the diurnal evolution of rainfall over different surface types near vicinity of Darwin according to the MJO phases (Figs. 14, 15). For brevity, only the 3-hourly centered averaged maps at every 4-h interval are shown. MJO phases 4–6 show both a common and a different type of temporal evolution in rainfall with varying amplitude depending upon the location (Fig. 14). For example, over the Beagle Gulf during these phases, the rainfall starts to develop around midnight (2200–2400 LT) which further strengthens and spreads out towards east during early morning (0200–0400 LT) and reaches its minimum during late evening (1800–2000 LT). However, the spreading of early morning intense rainfall towards coastal region of Darwin is extensive in phases 5 (see Fig. 13) and 6 while it is limited to near the coast in phase 4. Over the Tiwi Islands, during phases 4–6, rainfall starts to build up before noon (1200 LT), becomes prominent during afternoon (1400–1600 LT), and ceases during evening (1800–2000 LT). However, the strength of the diurnal cycle is weaker during these phases due to suppression of sea breezes causing about 50 % reduction in the diurnal amplitude compared to what is observed during the MJO suppressed phases (compare Figs. 14 and 15). Over the mainland during these phases, the diurnal cycle of rainfall shows a different temporal evolution. During phase 4, rainfall mainly occurs over the southeastern corner of the radar domain where it develops around afternoon (1400–1600 LT) and attains its maximum during late evening (1800– 2000 LT). In contrast, during phase 6, the rainfall mainly occurs over the southwestern corner of the radar domain
13
S. P. Rauniyar, K. J. E. Walsh
Fractional Variances (%)
18 16
1.00
(a) Eigenvalue Spectrum
(b) EOFs
●
14
0.50
12
0.25
●
10
0.00
8
−0.25
● ●
6
●
●
−0.50
● ●
4
●
01
02
03
04
05
06
07
08
●
09
−0.75
10
00
03
06
EOFs
Rain anomaly
50
EOF1 EOF2
0.75
09
12
15
18
21
24
Central standard time (hr)
(c) PC1
25 0 −25 −50
Rain anomaly
50
(d) PC2
25 0 −25 −50 02/03
2.5 2.0 1.5 1.0 0.5 0.0 −0.5 −1.0 −1.5
03/04
04/05
05/06
(e) MJO Phases
●
●
06/07
07/08
(f) Large−scale Regimes
PC1 PC2 ●
● ●
●
●
●
●
●
7
8
● ●
●
1
2
3
4
5
6
W
DW
SW
ME
Fig. 12 The results of EOF analysis of hourly rainfall averaged using the common 6 DJF (2002/03–2007/08) seasons of the BoM and the ACRF sites: a eigenvalue spectrum with error bars based on North et al. (1982) significance test, b time series of first two dominant
EOFs, c, d the time series of PC1 and PC2, e the composites of PC1 (filled circle) and PC2 (filled triangle) as per the MJO phases 1–8 and weak, and f the same as (e), but for the DW, SW and ME Darwin regimes
where it initiates along the coast during early afternoon (1000–1200 LT) and becomes prominent during late afternoon (1400–1600 LT). These spatial pattern in the diurnal cycles during phases 4 and 6 produce a distinct dipole patterns in rainfall anomalies as seen in Fig. 9 and may be associated with remote rainfall influences. During phase 4, the active center of large-scale convection is located over the Timor Sea northwest of Australia (WH04). This produces a suppressed large-scale condition over Cape York Peninsula in northern
Queensland. In contrast, during phase 6, the most active center of MJO related convection is generally situated over the northeast region of Australia (Cape York Peninsula and the Coral Sea). It has been found that the formation of long-lived westward propagating squall lines over Cape York Peninsula is enhanced (suppressed) when the largescale condition is suppressed (enhanced) over that region (Drosdowsky and Holland 1987; May et al. 2012). These squall lines propagate westward through the Gulf of Carpentaria and tend to dissipate as they approach the Darwin
13
Spatial and temporal variations in rainfall over Darwin and its vicinity during different…
Fig. 13 Upper panel shows the variance explained in percentage by harmonics 1 and 2 in the hourly mean diurnal cycle of rainfall for the MJO phase 5 while the lower panel shows their corresponding phases (time of maximum rainfall in LT)
region in the late evening due to the existence of a cooler surface layer as a result of earlier convection (May et al. 2012). Hence, we speculate that the above normal rainfall over southeast of the radar domain during phase 4 may be caused by these dissipating systems, which are suppressed during phase 6 resulting in below normal rainfall over the same region. However, we still do not understand the full dynamics behind the late and early initiation of rainfall during the phases 4 and 6, respectively over the southeastern and southwestern regions of the radar domain. During the suppressed phases (1–3 and 8) of the MJO, a prominent diurnal cycle in rainfall only exists over land and islands (Fig. 15). Over the Tiwi Islands, during these phases, the rainfall is mainly produced by the Hector which initiates just after noon (1200 LT), matures during late afternoon (1400–1600 LT), yielding the heaviest amount of rainfall compared to anywhere else in the radar domain, and then dissipates in the evening (1800 LT). The amplitude of afternoon rainfall during these phases shows that the strength of the Hectors is the weakest in phase 8, further strengthens in phase 1, and becomes strongest in phase 2, after which it weakens in phase 3. There are two major
modes of Hector development over the Tiwi Islands (Carbone et al. 2000; Crook 2001). These are convergence of sea breezes from all coastlines, and the interaction of sea breezes with cold pools generated from earlier convection. According to Carbone et al. (2000), the second mechanism produces the heaviest rainfall over the Tiwi Islands. Crook (2001) showed that a higher value of low-level (below 1500 m) moisture is required for the generation of evaporatively produced cold pools in their non-linear modeling. However, we have not found such a distinction in low level moisture between MJO phases 2 and 3 (Fig. 5b). Hence, a further investigation using a high-resolution numerical model is required to understand which mechanism plays the more important role in producing the heaviest rainfall in phase 2. Over the mainland, during the suppressed MJO phases, the rainfalls are mainly produced by isolated afternoon convection, but they have different life cycles. In phase 2, the maximum rainfall over the mainland occurs almost during the same time (1400–1600 LT) as over the Tiwi Islands due to early initiation of sea breezes in the afternoon. In contrast, in phases 8 and 1, it is delayed until evening
13
S. P. Rauniyar, K. J. E. Walsh
Fig. 14 Diurnal evolution of the CPOL rainfall (mm h−1) at every 4 h interval for the MJO phases 4–7. Each map represents the centered average of 3 consecutive hours which has been smoothed in the X- and Y-directions by applying a 5-point boxcar filter
13
Spatial and temporal variations in rainfall over Darwin and its vicinity during different…
Fig. 15 The same as in Fig. 14, but for the MJO phases 8, and 1–3
13
S. P. Rauniyar, K. J. E. Walsh
(1800–2000 LT) and mostly occurs along the southeastern edge of the radar domain due to late initiation and inland penetration of sea breezes. Over the Van Diemen Gulf during the suppressed phases, the maximum rainfall is observed during early morning (0200–0400 LT) and has larger amplitude compared to the Beagle Gulf. This is due to the convergence of two land breezes, one originating from mainland and the other from the Tiwi Islands (Wapler and Lane 2012) or may also be caused by interactions of prevailing easterly flows with a single land breeze (Keenan et al. 1988). We also compare the diurnal characteristics of rainfall during the most active and suppressed phases of the MJO with the active and suppressed conditions of monsoon to identify any differences between them. Although the deep and moist westerly winds prevail with widespread rainfall during the active monsoon periods (DW regime) and also during the most active MJO periods (phase 5), the characteristics of diurnal cycle of rainfall differ significantly (compare second column of Fig. 15 with first column of Fig. 16). For example, over the Beagle Gulf, in phase 5, the maximum rainfall is observed during early morning (0200–0400 LT) whereas it is delayed till morning (0600– 0800 LT) in the DW regime. Similarly, the southwestern coastal region near Darwin receives more rainfall during late afternoon (1400–1600 LT) in phase 5 whereas the same region experiences the maximum rainfall during early afternoon (1000–1200 LT). This explains the occurrence of two consecutive afternoon peaks at 1200 LT and 1600 LT in EOF1 (Fig. 12b). In addition, over the southeastern region of radar domain, the rainfall peaks during late afternoon (1400–1600 LT) in the DW regime while in phase 5, the same peak is delayed and occurs during evening (1800–2000 LT). Over the Tiwi Islands, the area of maximum rainfall is located near the eastern coast in the DW regime whereas it is located over the western part in phase 5. This is possible as phase 5 shows the combined effect of ME, DW and SW regimes as they contribute almost equally (Fig. 4c). In contrast, the diurnal rainfall characteristics between the ME regime and phase 2 has very similar features as the frequency of the ME regime is much higher in phase 2 compared to other regimes.
6 Discussion and summary This paper provides climatological characteristics of Australian summer (DJF) rainfall over various surface types (land, island, coastal and ocean) near Darwin, Australia under the influence of different synoptic environments by utilizing a combination of in situ and CPOL radar data that is longer term and higher resolution than any of the previous studies in this region. Although many indices exist
13
to study the rainfall variability over Darwin region, two approaches are now more common: the five Darwin atmospheric regimes as identified by Pope et al. (2009a) and the eight strong phases of MJO as proposed by WH04. However, we found that statistical inferences of rainfall variability using the Darwin regimes with only a few seasons of data have significant limitations as these regimes vary considerably during alternate phases of ENSO. In contrast, the MJO phases, which show the location of large-scale regions of enhanced and suppressed convection, have insignificant inter-annual variation suggesting that the rainfall climatology according to these phases can be developed with fewer seasons of data. This is expected as the MJO phases are determined using the large-scale horizontal winds and OLR fields, whereas the regimes are identified using local radiosonde data from Darwin. Furthermore, the propagating nature of the MJO enables us to construct life cycles of rainfall-building mechanisms which are not possible using the regimes due to their non-regular pattern of occurrence (propagation). Hence, most of the analysis in this study were done using the MJO phases, except as otherwise stated. The results shown in this study support previous work that indicates a large spatial and temporal variability in rainfall over the study domain, which is mainly controlled by two mechanisms: the large-scale and sea breeze mechanisms. The large-scale mechanism represents the condition when the Darwin region experiences westerly winds at lowto mid-level with a wetter atmosphere than climatology and produces widespread rainfall over the entire study domain. These features are found to be prevalent only during the MJO phases 4–7. More specifically, the large-scale condition starts to build up from MJO phase 4 with the reversal of lower tropospheric dry easterly winds into weak, but moist westerly winds. This is caused by the arrival of MJO’s convective envelop over the Timor Sea, which inhibits the local-scale sea breezes over the study domain by reducing the local-scale land–sea thermal contrast significantly. During phase 5, the large-scale mechanism attains its peak due to the actual arrival of convectively active MJO over the top end of Australia. This strengthens the low- to mid-level westerly winds making the atmosphere the wettest of all phases, but reduces the land surface temperature significantly compared to local SST. As the active centre of MJO moves toward the Coral Sea in phase 6, the effect of the large-scale mechanism starts to decline over Darwin as seen in the weakening of mid-level (2–7 km) westerly winds and in the reduction of mid-atmospheric moisture. In phase 7, the strength of the large-scale mechanism further reduces due to the turning of mid-level westerly winds into southerly winds, which bring dry air from the continent resulting in a significant decrease in atmospheric moisture. Furthermore, the existence of shallow (below 2 km) but
Spatial and temporal variations in rainfall over Darwin and its vicinity during different…
Fig. 16 The same as in Fig. 14, but for the Darwin DW (active monsoon) and ME (break monsoon) regimes
strong westerly winds suppresses the initiation of any sea breezes resulting into a significant decrease in rainfall over the entire study domain. In general, the region experiences an increase in mean rainfall during phases 4–6, and this is not a new result. However, the present study shows that remarkable spatial and temporal anomalies exist in the observed rainfall distribution. In phase 4, above normal rainfall is found to be located over the southeastern corner of the radar domain. The diurnal composite showed that the maximum rainfall over the southeastern region occurs during late evening (1800–2000 LT). In contrast, in phase 6, the positive rainfall anomaly exists over the southwestern corner of the radar domain and maximum rainfall is found to be in late afternoon (1400–1600 LT), which is 2–4 h earlier than what is observed in phase 4. These anomalies do not seem to be due to the local sea breezes as no such amplification in afternoon rainfall over the Tiwi Islands is observed either in phase 4 or in phase 6. A possible alternate explanation may be related to the enhanced and suppressed conditions of the long-lived westward propagating squall lines that originate from Cape York Peninsula during phase 4 and 6, respectively when the large-scale condition is suppressed and enhanced over there, but analysis has yet not be performed to demonstrate this relationship. In phase 5, the heavy and widespread rainfall is observed over the entire study domain, but the most intense rainfall which occurs during early morning (0200–0400 LT) is found to be along the southwestern coast of Darwin in the Beagle Gulf which explains the occurrence of early morning (0200–0400 LT) peak in the diurnal cycle of Darwin rainfall. These results show that the maximum rainfall (phase 5) in this region leads the strongest westerly winds (phase 7) by almost two MJO phases (10–12 days), which is consistent with Wheeler and McBride (2011). During the sea breeze mechanism dominated period, the Darwin region experiences low- to mid-level dry easterly/ southeasterly winds with a significantly increase in the afternoon near surface temperature at Darwin compared to the adjacent SST and are found to be prevalent during the MJO phases 8 and 1–3. More specifically, the reinvigoration of sea breezes takes place during phase 8 probably due to a significant decrease in the strength of shallow westerly winds. However, the existence of stronger mid-tropospheric southeasterly winds makes the atmosphere the driest of all phases. As a result, rainfall is predominately concentrated over the Tiwi Islands and also along the southwestern coast of mainland and occurs mostly during late afternoon (1400–1600 LT). The strength of sea breezes further increases in phase 1 and reaches maximum in phase 2 due to the strengthening of low to mid-level southeasterly/easterly winds and increase in mid-atmospheric moisture. As a
13
S. P. Rauniyar, K. J. E. Walsh
result, the largest amount of rainfall is observed in phase 2 over the Tiwi Islands during late afternoon (1400–1600 LT) compared to anywhere else in the radar domain. This suggests that the Hectors, which are the main contributor of rainfall over the Tiwi Islands, are more frequent during phase 2. However, a further investigation is needed to explore the mechanism behind higher occurrence of Hectors during phase 2. In phase 3, the magnitude of lower level easterly winds starts to reduce over Darwin making the region still favourable for stronger sea breezes. In contrast, the decrease in afternoon rainfall over the Tiwi Islands in phase 3 suggests weakening of the sea breezes over the Tiwi Islands. Nevertheless, during MJO phases 8 and 1–3, the amplitudes of afternoon rainfall over the Tiwi Islands are more than twice the amplitudes observed during MJO phases 4–7. EOF analysis applied on in situ hourly rainfall data to explore the mechanism behind the occurrence of the early morning (0200–0400 LT) peak in the diurnal cycle of rainfall at Darwin showed two dominant non-degenerate diurnal patterns which together explain more than 27 % of total variance. The early morning rainfall peak is found to be most significant in EOF1 and is mainly associated with the MJO phases 4–6. Harmonic analysis of CPOL rainfall showed that the maximum rainfall during the early morning also occurs over southwestern coastal region near Darwin. The phase distribution of second harmonics (semi-diurnal cycles) composited for the MJO phases 4–6 showed that the early morning rainfall over these regions is caused by an eastward advection or expansion of rainfall systems that form over the Beagle Gulf during mid-night (0000–0200 LT). In contrast, the EOF2 which explains 12 % of total variance has exhibited a strong diurnal peak in the afternoon (1500–1700 LT) and found to be associated with the MJO phases 1–3 and 8 or the ME regime. Furthermore, a comparison of the diurnal evolution of rainfall during the active monsoon period (DW regime) with the most active MJO phase 5 showed significantly different diurnal characteristics despite having very similar synoptic conditions in wind and humidity profiles (i.e., deep and moist westerly winds). In the DW regime, the maximum rainfall over the Beagle Gulf is observed at morning (0600–0800 LT), which is a delay of approximately 2 to 4 h compared to phase 5. In contrast, over the southwestern coastal region near Darwin, the daytime rainfall peaks during early afternoon (1000–1200 LT) in the DW regime compared to late afternoon (1400–1600 LT) in phase 5. Although the maximum rainfall occurs almost at the same time over the Tiwi Islands, the area of maximum rainfall is located near the eastern coast in the DW regime whereas the same is located over the western part in phase 5. These anomalies are due to a slightly higher contribution of the ME regime in phase 5. In contrast, the
13
diurnal rainfall characteristics during the break monsoon period (ME regime) has a very similar feature as observed in the MJO suppressed phase 2 as the contribution of the ME regime is much higher in phase 2 compared to other regimes. In summary, our main conclusions are: • There are substantial regional variations in the spatial and temporal development of rainfall in the vicinity of Darwin with the changing phases of MJO. • The bimodal peaks in the diurnal rainfall in the region only occurs in coastal areas near Darwin during the MJO active phase due to advection or expansion of rainfall system from the Beagle Gulf. • The diurnal variation of rainfall differs between the MJO-defined active phase and the monsoon-defined active phase despite the similarity between their vertical wind and moisture profiles. Acknowledgments The authors wish to thank Dr. Matthew Wheeler, Dr. Peter May and Dr. Alain Protat for their valuable suggestions and comments. We would also like to thank Michael Whimpey for providing the CPOL radar data. In this study, ERA-Interim data were obtained from the European Centre for Medium-Range Weather Forecasts and atmospheric variables datasets were obtained from the ARM Climate Modeling Best Estimate (CMBE) atmospheric and radiation datasets and from the Australian Bureau of Meteorology.
References Arritt RW (1993) Effects of the large-scale flow on characteristic features of the sea breeze. J Appl Meteorol 32(1):116–125 Beringer J, Tapper NJ, Keenan TD (2001) Evolution of maritime continent thunderstorms under varying meteorological conditions over the Tiwi Islands. Int J Climatol 21(8):1021–1036 Bringi V, Huang G-J, Chandrasekar V, Keenan T (2001) An areal rainfall estimator using differential propagation phase: evaluation using a C-band radar and a dense gauge network in the tropics. J Atmos Ocean Technol 18(11):1810–1818 Bringi V, Tang T, Chandrasekar V (2004) Evaluation of a new polarimetrically based Z-R relation. J Atmos Ocean Technol 21(4):612–623 Caine S, Jakob C, Siems S, May P (2009) Objective classification of precipitating convective regimes using a weather radar in Darwin, Australia. Mon Weather Rev 137(5):1585–1600 Carbone R, Wilson J, Keenan T, Hacker J (2000) Tropical island convection in the absence of significant topography. Part I: Life cycle of diurnally forced convection. Mon Weather Rev 128(10):3459–3480 Crook NA (2001) Understanding Hector: the dynamics of island thunderstorms. Mon Weather Rev 129(6):1550–1563 Crosman ET, Horel JD (2010) Sea and lake breezes: a review of numerical studies. Bound-Layer Meteorol 137(1):1–29 Drosdowsky W (1996) Variability of the Australian summer monsoon at Darwin: 1957–1992. J Clim 9(1):85–96 Drosdowsky W, Holland G (1987) North australian cloud lines. Mon Weather Rev 115(11):2645–2659 Grimsdell AW, Alexander MJ, May PT, Hoffmann L (2010) Model study of waves generated by convection with direct validation via satellite. J Atmos Sci 67(5):1617–1631
Spatial and temporal variations in rainfall over Darwin and its vicinity during different… Hamilton K, Vincent RA, May PT (2004) Darwin Area Wave Experiment (DAWEX) field campaign to study gravity wave generation and propagation. J Geophys Res 109:1–15 Hendon HH, Liebmann B (1990a) A composite study of onset of the Australian summer monsoon. J Atmos Sci 47(18):2227–2240 Hendon HH, Liebmann B (1990b) The intraseasonal (30–50 day) oscillation of the Australian summer monsoon. J Atmos Sci 47(24):2909–2924 Holland GJ, McBride JL, Smith RK, Jasper D, Keenan TD (1986) The BMRC Australian monsoon experiment: AMEX. Bull Am Meteorol Soc 67(12):1466–1472 Hung CW, Yanai M (2004) Factors contributing to the onset of the Australian summer monsoon. Q J R Meteorol Soc 130(597):739–758 Kanamori H, Yasunari T, Kuraji K (2013) Modulation of the diurnal cycle of rainfall associated with the MJO observed by a dense hourly rain gauge network at Sarawak, Borneo. J Clim 26(13):4858–4875 Keenan TD, Carbone RE (1992) A preliminary morphology of precipitation systems in tropical northern Australia. Q J R Meteorol Soc 118(504):283–326 Keenan TD, Holland GJ, Manton MJ, Simpson J (1988) TRMM ground truth in a monsoon environment: Darwin, Australia. Aust Meteorl Magazine 36:81–90 Keenan TD, Manton MJ, Holland GJ, Morton BR (1989) The Island Thunderstorm Experiment (ITEX)—a study of tropical thunderstorms in the maritime continent. Bull Am Meteorol Soc 70(2):152–159 Keenan T, Morton B, Zhang XS, Nyguen K (1990) Some characteristics of thunderstorms over Bathurst and Melville Islands near Darwin, Australia. Q J R Meteorol Soc 116(495):1153–1172 Keenan T, Glasson K, Cummings F, Bird T, Keeler J, Lutz J (1998) The BMRC/NCAR C-band polarimetric (C-Pol) radar system. J Atmos Ocean Technol 15(4):871–886 Keenan T, May P, Holland G, Rutledge S, Carbone R, Wilson J, Moncrieff M, Crook A, Takahashi T, Tapper N (2000) The maritime continent thunderstorm experiment (MCTEX): overview and some results. Bull Am Meteorol Soc 81(10):2433–2455 Koike T (2004) The Coordinated Enhanced Observing Period—an initial step for integrated global water cycle observation. WMO Bull 53(2):115–121 Kumar VV, Protat A, May PT, Jakob C, Penide G, Kumar S, Davies L (2013) On the effects of large-scale environment and surface types on convective cloud characteristics over Darwin, Australia. Mon Weather Rev 141(4):1358–1374 Madden RA, Julian PR (1994) Observations of the 40–50-day tropical oscillation—a review. Mon Weather Rev 122(5):814–837 May PT, Ballinger A (2007) The statistical characteristics of convective cells in a monsoon regime (Darwin, Northern Australia). Mon Weather Rev 135(1):82–92 May PT, Mather JH, Vaughan G, Bower KN, Jakob C, McFarquhar GM, Mace GG (2008) The tropical warm pool international cloud experiment. Bull Am Meteorol Soc 89(5):629–645 May PT, Long CN, Protat A (2012) The diurnal cycle of the boundary layer, convection, clouds, and surface radiation in a coastal monsoon environment (Darwin, Australia). J Clim 25(15):5309–5326 McBride JL, Frank WM (1999) Relationships between stability and monsoon convection. J Atmos Sci 56(1):24–36 Miller S, Keim B, Talbot R, Mao H (2003) Sea breeze: structure, forecasting, and impacts. Rev Geophys 41(3):1–31 North GR, Bell TL, Cahalan RF, Moeng FJ (1982) Sampling errors in the estimation of empirical orthogonal functions. Mon Weather Rev 110(7):699–706 Oliphant A, Sturman A, Tapper N (2001) The evolution and structure of a tropical island sea/land-breeze system, northern Australia. Meteorol Atmos Phys 78(1–2):45–59
Pope M, Jakob C, Reeder MJ (2009a) Regimes of the north Australian wet season. J Clim 22(24):6699–6715 Pope M, Jakob C, Reeder MJ (2009b) Objective classification of tropical mesoscale convective systems. J Clim 22(22):5797–5808 Porson A, Steyn DG, Schayes G (2007) Formulation of an index for sea breezes in opposing winds. J Appl Meteorol Climatol 46(8):1257–1263 Protat A, Delanoë J, May P, Haynes J, Jakob C, O’Connor E, Pope M, Wheeler M (2011) The variability of tropical ice cloud properties as a function of the large-scale context from ground-based radar-lidar observations over Darwin, Australia. Atmos Chem Phys 11:8363–8384 Protat A, Rauniyar S, Kumar VV, Strapp JW (2014) Optimizing the probability of flying in high ice water content conditions in the tropics using a regional-scale climatology of convective cell properties. J Appl Meteor Climatol 53(11):2438–2456 Ramage CS (1968) Role of a tropical maritime continent in the atmospheric circulation. Mon Weather Rev 96(6):365–370 Rauniyar SP (2006) Characteristics and mechanism analysis of precipitation’s diurnal cycle based on CEOP. Dissertation (M. Sc.), University of Tokyo, Japan Rauniyar SP, Rasmy M, Tamagawa K, Yang K, Koike T (2007) Prediction skill assessment of NWP models in simulating the diurnal cycle of precipitation. Annu J Hydraulic Eng, JSCE 51:97– 102. doi:10.2208/prohe.51.97 Reynolds RW, Smith TM, Liu C, Chelton DB, Casey KS, Schlax MG (2007) Daily high-resolution-blended analyses for sea surface temperature. J Clim 20(22):5473–5496 Santos e Silva CM, Gielow R, de Freitas SR (2009) Diurnal and semidiurnal rainfall cycles during the rain season in SW Amazonia, observed via rain gauges and estimated using S-band radar. Atmos Sci Lett 10(2):87–93 Skinner T, Tapper N (1994) Preliminary sea breeze studies over Bathurst and Melville Islands, Northern Australia, as part of the Island Thunderstorm Experiment (ITEX). Meteorol Atmos Phys 53(1–2):77–94 Soman VV, Valdés JB, North GR (1995) Satellite sampling and the diurnal cycle statistics of Darwin rainfall data. J Appl Meteorol 34(11):2481–2490 Troup A (1961) Variations in upper tropospheric flow associated with the onset of the Australian summer monsoon. Indian J Meteorol Geophys 12:217–230 Wapler K, Lane TP (2012) A case of offshore convective initiation by interacting land breezes near Darwin, Australia. Meteorl Atmos Phys 115:123–137 Wheeler MC, Hendon HH (2004) An all-season real-time multivariate MJO index: development of an index for monitoring and prediction. Mon Weather Rev 132(8):1917–1932 Wheeler M, McBride J (2011) Australasian monsoon. In: Lau WKM, Waliser D (eds) Intraseasonal variability in the atmosphere–ocean climate system, 2nd edn. Springer, New York, pp 147–198 Wheeler MC, Hendon HH, Cleland S, Meinke H, Donald A (2009) Impacts of the Madden–Julian oscillation on Australian rainfall and circulation. J Clim 22(6):1482–1498 Xie S, McCoy RB, Klein SA, Cederwall RT, Wiscombe WJ, Jensen MP, Johnson KL, Clothiaux EE, Gaustad KL, Long CN, Mather JH, McFarlane SA, Shi Y, Golaz J-C, Lin Y, Hall SD, McCord RA, Palanisamy G, Turner DD (2010) CLOUDS AND MORE: ARM climate modeling best estimate data. Bull Am Meteorol Soc 91(1):13–20 Yang S, Smith EA (2006) Mechanisms for diurnal variability of global tropical rainfall observed from TRMM. J Clim 19(20):5190–5226 Zhang C (2005) Madden–Julian oscillation. Rev Geophys 43:RG2003. doi:10.1029/2004RG000158
13