Abstract

Straddling the Asian–Australian monsoon region, the Maritime Continent (MC) experiences substantial rainfall variations from diurnal to interannual and longer time scales. In this study, rainfall over Singapore and the wider MC region are analyzed using objectively identified weather regimes. Eight regional-scale weather regimes are derived by k-means clustering of local vertical profiles of zonal and meridional winds, temperature, and specific humidity extracted over Singapore from ERA-Interim data for the period December 1980–November 2014. The composite synoptic flow and rainfall patterns over the region show that the weather regimes correspond to the seasonal migration of the intertropical convergence zone (ITCZ) across the equator. For Singapore, the regimes depict seasonal rainfall variability by capturing the alternating dry and wet phases of the prevailing local monsoon and transition periods associated with the regional-scale ITCZ movement. Following previous work, the regimes are used to examine the annual rainfall trend by calculating the contributions due to 1) changes in regime frequency, indicating regional-scale circulation changes, and 2) changes in within-regime precipitation, indicating altered thermodynamic conditions. The overall trend observed at Singapore and many other MC locations is overwhelmingly due to changes in within-regime precipitation. However, the overall trend masks the larger contribution resulting from regime frequency changes as these circulation changes tend to offset one another in reality. In many MC areas (including Singapore), summed rainfall changes due to regime frequency changes outweigh those due to changes in within-regime rainfall, when aggregated in an absolute sense.

1. Introduction

The Maritime Continent (MC) is one of the warmest and wettest regions on Earth. Situated within the tropical warm pool between the Indian and Pacific Oceans, the Maritime Continent separates mainland Asia from Australia with its unique geography of landmasses surrounded by very warm (≥28°C) seas. The region itself comprises several large mountainous islands (Sumatra, Kalimantan, Java, Sulawesi, and New Guinea), the Malay Peninsula, the Philippines, and thousands of other smaller islands that constitute the rest of the Indonesian archipelago.

Climatologically, the Maritime Continent is characterized by a very heterogeneous distribution of annual mean rainfall and exhibits substantial variability across the region on seasonal time scales, particularly over land and coastal areas (Fig. 1) based on data from the Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA) 3B42 product (Huffman et al. 2007, 2010). To first order, such a diverse spatial distribution is arguably governed by the close interaction between a strong topographically induced diurnal cycle (Yang and Slingo 2001; Qian 2008; Kikuchi and Wang 2008; Yamanaka 2016) and the large-scale seasonal flow associated with the Asian–Maritime Continent–Australian monsoon system (Chang et al. 2005b; Robertson et al. 2011).

Fig. 1.

Long-term (top) annual and (middle),(bottom) seasonal (DJF, JJA, MAM, SON) mean daily rainfall (in mm day−1) from the TRMM Multisatellite Precipitation Analysis (TMPA) 3B42 product for the period December 1998–November 2014. Singapore is centrally located within the region denoted by the red box, at the tip of the Malay Peninsula in the western half of the Maritime Continent. The mean rainfall averaged within the red box is shown by the text in red.

Fig. 1.

Long-term (top) annual and (middle),(bottom) seasonal (DJF, JJA, MAM, SON) mean daily rainfall (in mm day−1) from the TRMM Multisatellite Precipitation Analysis (TMPA) 3B42 product for the period December 1998–November 2014. Singapore is centrally located within the region denoted by the red box, at the tip of the Malay Peninsula in the western half of the Maritime Continent. The mean rainfall averaged within the red box is shown by the text in red.

This is on top of the variability associated with synoptic-scale disturbances arising from the passage of equatorial waves (Liebmann and Hendon 1990) and from seasonal phenomena such as monsoon cold surges (Lim et al. 2017), Borneo vortices, and other South China Sea vortices (Chang et al. 2005a; Tangang et al. 2008; Chen et al. 2015; Nguyen et al. 2016) during boreal winter. The Madden–Julian oscillation, being the dominant mode of intraseasonal variability, is also known to systematically modulate the diurnal cycle and daily rainfall over the Maritime Continent (Rauniyar and Walsh 2011; Oh et al. 2012; Peatman et al. 2014; Birch et al. 2016; Vincent and Lane 2016). In addition, several studies have shown that ENSO exerts a substantial and varying impact on rainfall across many areas in the Maritime Continent on interannual time scales. For example, Haylock and McBride (2001) and McBride et al. (2003) noted that rainfall across the entire Indonesian region was highly correlated with ENSO during the dry season and transition periods but has little correlation during the peak of the wet season in December and January. Conversely, Chang et al. (2004) demonstrated that interannual rainfall during the wet season behaves differently in subregions across the Maritime Continent. They showed that while rainfall over the Sumatra–Malay Peninsula sector has a weak positive correlation with Niño-3 sea surface temperatures (SSTs), surrounding areas in the central Maritime Continent and the oceanic region west of Sumatra actually display significant negative SST–rainfall correlations for the same season. Chang et al. (2004) further noted that western and eastern Indonesia have opposite rainfall responses to ENSO phases during the wet season and attributed the low correlation found in earlier studies to be partly due to the averaging of rainfall across the two subregions.

Singapore (1.35°N, 103.8°E) is situated at the tip of the Malay Peninsula in the western half of the Maritime Continent (see red box of Fig. 1). Like other nearby equatorial locations across the region, it receives a considerable amount of rain throughout the year (~2493 mm yr−1, based on a 28-station average between 1981 and 2010) and experiences two main monsoonal periods that are generally defined by the seasonal reversal of the mean prevailing low-level wind direction over Singapore at 850 hPa and below. The wetter boreal winter monsoon, locally known as the northeast (NE) monsoon, typically occurs between December and April, while the relatively drier boreal summer monsoon happens during June to September and is termed the southwest (SW) monsoon. The two local monsoon seasons are interspersed by the shorter transitional (intermonsoon) periods of April–May and October–November.

Singapore has no distinct wet or dry season, yet the mean annual cycle of precipitation shows considerable variability (Fig. 2). The largest month-to-month rainfall variation is embedded within the NE monsoon, with December receiving about 50% more rainfall than February. In contrast, long-term monthly totals during the SW monsoon are relatively steadier. Singapore has also experienced a significant upward trend in annual total rainfall over the 1981–2014 period, as derived from station gauge data (Fig. 3). The trend is consistent with that obtained from a 2.5° grid box centered on Singapore from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) 2.5° × 2.5° pentad product (Xie and Arkin 1997). Notably, large areas within the Maritime Continent also experienced significant upward trends according to CMAP for the same period (Fig. 4).

Fig. 2.

Mean monthly rainfall over Singapore for the period December 1980–November 2014, as averaged across 28 stations island-wide (blue line; vertical axis on left), and from a corresponding CMAP grid point (blue dashed line). Linear trends in station-averaged (filled bars) and CMAP monthly rainfall are shown with values (vertical axis on right; mm yr−1). Statistically significant monthly trends at the 5% level are highlighted by the red outline.

Fig. 2.

Mean monthly rainfall over Singapore for the period December 1980–November 2014, as averaged across 28 stations island-wide (blue line; vertical axis on left), and from a corresponding CMAP grid point (blue dashed line). Linear trends in station-averaged (filled bars) and CMAP monthly rainfall are shown with values (vertical axis on right; mm yr−1). Statistically significant monthly trends at the 5% level are highlighted by the red outline.

Fig. 3.

December–November accumulated rainfall totals over Singapore (December 1980–November 2014) from rain gauge data (28-station average, black) and a corresponding CMAP 2.5° × 2.5° grid point (blue). Linear trends shown (dashed lines) are estimated using the nonparametric Theil–Sen slope estimator. A significant monotonic trend is detected if the associated probability P from the Mann–Kendall test is P . The cross-correlation coefficient between gauge data and CMAP is 0.89.

Fig. 3.

December–November accumulated rainfall totals over Singapore (December 1980–November 2014) from rain gauge data (28-station average, black) and a corresponding CMAP 2.5° × 2.5° grid point (blue). Linear trends shown (dashed lines) are estimated using the nonparametric Theil–Sen slope estimator. A significant monotonic trend is detected if the associated probability P from the Mann–Kendall test is P . The cross-correlation coefficient between gauge data and CMAP is 0.89.

Fig. 4.

Linear trends in annual (December–November) accumulated rainfall in mm yr−1 for the Maritime Continent. Hatched regions are areas where the associated probability P of a significant monotonic trend is according to the Mann–Kendall test.

Fig. 4.

Linear trends in annual (December–November) accumulated rainfall in mm yr−1 for the Maritime Continent. Hatched regions are areas where the associated probability P of a significant monotonic trend is according to the Mann–Kendall test.

A couple of key questions therefore arise: 1) What determines the monthly-to-seasonal variability? 2) What drives the observed increase in annual Singapore rainfall since 1981? Thus, our goal is to seek a mechanistic understanding of the long-term rainfall trends, regardless of their magnitudes and statistical significance. These measures can and do change depending on the start and end dates of the analysis. For example, extending the analysis for CMAP rainfall to 2017 revealed different trend values and loss of statistical significance over many areas. Nonetheless, the spatial patterns in the sign of the annual rainfall trend remain robust (not shown).

To shed light, a promising approach is the concept of regional-scale weather regimes. The notion of weather regimes can be thought of as a useful way to organize recurrent meteorological states of the atmosphere that make the climate of a region. The idea of classifying the atmosphere into states stems from the observation that local weather patterns are considerably related to specific structures of the regional-scale flow regime. Each weather state is therefore characterized by distinct mean spatial patterns of wind, moisture, rainfall, and temperature.

Classifying the broad-scale circulation through weather regimes also has the advantage of providing information on the physical processes of the atmosphere governing regional climate change (Hertig and Jacobeit 2014). Hence, it can be used as a tool to understand day-to-day weather as well as long-term climate variability and change. Furthermore, it can be a method to evaluate model ability and utilized as a means to downscale large-scale climate change information when regimes are identified and linked to particular local conditions. Since the pioneering work of Hess and Brezowsky (1952) and Lamb (1972), who manually classified synoptic patterns to describe seasonal variability over Europe and the United Kingdom, respectively, weather regimes and cluster analysis have been used extensively to diagnose key large-scale atmospheric states and examine their relationship to regional and local weather over Europe and North America (Vautard and Legras 1988; Vautard 1990; Michelangeli et al. 1995; Robertson and Ghil 1999; Yiou and Nogaj 2004; Hertig and Jacobeit 2014).

In the tropics, studies are more recent and weather typing has been used to describe wet season variability over East Africa (Pohl et al. 2005), Senegal (Moron et al. 2008), Darwin (Pope et al. 2009), and Indonesia (Moron et al. 2010; they also looked at the coherence and seasonal predictability of monsoon onset). Specific studies on the influence of ENSO on diurnal rainfall variability have also been performed for Java (Qian et al. 2010) and Borneo (Qian et al. 2013) through its impact on associated weather types, using the same number of weather types (five for the August–February season), as defined by Moron et al. (2010). Recently, Moron et al. (2015) defined six weather types using the same technique for the entire Maritime Continent when they analyzed an extended September–April period to cover the full austral summer. Except for Pope et al. (2009), who used radiosonde profiles over Darwin, all of the other aforementioned studies defined the wet season weather types using low-level regional wind and/or outgoing longwave radiation (OLR) fields. None of the above tropical studies have defined weather types for the whole year. Doing so would help to improve the understanding of regional mechanisms driving the annual cycle of rainfall for more equatorial locations such as Singapore.

Given that Singapore experiences rain throughout the year and has no “dry” season, there is a need to objectively define weather regimes that cover its entire annual cycle. In this study, we adapted the k-means clustering technique used by Pope et al. (2009) and applied it to pressure-level ERA-Interim data extracted over Singapore (section 2). We identify and characterize eight principal weather regimes that influence not only Singapore’s annual rainfall distribution (section 3) but also the wider Maritime Continent, thus providing a unified picture for the entire annual cycle of rainfall across the whole basin north to south (section 4). We interpret the regional-scale fundamental weather states as corresponding to the seasonal migration of the intertropical convergence zone (ITCZ) across the equator. Our work extends previous analyses by including the boreal summer season to complete the full 12-month cycle, while characterizing boreal summer variability by identifying the associated regimes within. This is especially important since Lestari et al. (2015) show that projected drying signals in future boreal summers are significant and robust for the western Maritime Continent. Understanding what the regimes are for that season is a crucial first step to contextualizing the projected climate change signal. We also examine observed rainfall trends over Singapore and the whole Maritime Continent in terms of the relative frequency of occurrence and within-regime precipitation of the weather regimes (section 5). Section 6 presents a discussion and our conclusions.

2. Methods and data

a. k-means clustering

The k-means clustering technique was used to objectively derive the regional-scale regimes influencing Singapore rainfall. We utilized the “kmeans_as136” algorithm of Hartigan and Wong (1979) as implemented by the NCAR Command Language (http://dx.doi.org/10.5065/D6WD3XH5). The process of k-means clustering allows large amounts of data to be compressed into a set of localized concentrations. It uses an iterative algorithm to group X data elements around centroids for a specified number of cluster seeds, K. This is done by minimizing the within-cluster sum of squares (WCSS) for each cluster. The WCSS is simply the sum of the squared Euclidean distance between each element X in cluster , and the cluster centroid as given by

 
formula

We adopted the technique applied by Pope et al. (2009), who used daily radiosonde data to objectively define wet season (September–April) regimes over Darwin, northern Australia. However, quality checks on the Singapore radiosonde dataset found several points of discontinuity at various pressure levels due to major instrumentation changes at different points in time and missing data. To circumvent potential regime misclassifications due to data heterogeneity, we use the European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-Interim, herein ERAI; Dee et al. 2011) data instead. Reanalysis data consist of an optimal blend of observations and model hindcasts to represent our “best estimate” of the atmospheric state every 6 h. Since the modeling and data assimilation systems used are unchanged, the dataset produced is physically and dynamically consistent over the period being reanalyzed, making the interpretation of derived weather regimes more robust. The use of reanalysis data also means that this technique can be applied to locations that do not have radiosonde observations readily available.

To mimic radiosonde observations, we extracted profiles of air temperature, specific humidity, and zonal and meridional wind components from a single 0.125° × 0.125° grid box over Singapore at 1.375°N, 103.875°E. The 0.125° × 0.125° data are available from ECMWF as a download option, with the data being interpolated from ERAI’s native 0.75° × 0.75° model resolution. Daily data at 0000 UTC (0800 local time) were utilized for the period December 1980–November 2014 (12 418 days) from 12 pressure levels: 1000, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, and 100 hPa. The time 0000 UTC was chosen since land convection is generally at a minimum at this time, allowing us to sample environmental conditions not strongly modified by the presence of local convective systems. The calendar year is defined from December to November, following the traditional seasons of boreal winter (December–February), spring (March–May), summer (June–August), and autumn (September–November).

Prior to clustering, each variable is standardized at every level to remove the influence of magnitude due to the different units involved. The standardization ensures that equal weighting is given to each variable when clustering is done. The optimal number of clusters was determined by empirically increasing the value of K (), until only small reductions or an increase in the “total-distance variance” (TDV) are seen with larger K values. This would indicate that a stable number of clusters has been achieved. The change in TDV, following Jiang et al. (2016), is defined as

 
formula

where . Using this method, continued reduction in TDV is seen until when an increase in TDV is first registered for (Fig. 5): therefore, eight weather regimes were identified.

Fig. 5.

Change in total-distance variance (TDV). Only small reductions in TDV are seen with increasing number of predefined cluster seeds in the shaded area, indicating the stability of the cluster results at .

Fig. 5.

Change in total-distance variance (TDV). Only small reductions in TDV are seen with increasing number of predefined cluster seeds in the shaded area, indicating the stability of the cluster results at .

b. Rainfall and low-level wind field climatology

To create the mean rainfall climatology for each regime, we composite daily precipitation data derived from version 7 of the TMPA 3B42 dataset (Huffman et al. 2007, 2010) for corresponding cluster days. The TMPA 3B42 data are gauge-adjusted, multisatellite-derived estimates of precipitation obtained from passive microwave and infrared sensors. The data are 3-hourly and gridded with a spatial resolution of 0.25° × 0.25° between 50°N and 50°S. Despite its shorter record (1998–present), TMPA 3B42 data are favored over other satellite-based rainfall products for the Maritime Continent since it has the best agreement with gauge observations on daily and monthly time scales (Rauniyar et al. 2017). To analyze rainfall trends over the whole region, we utilize the CMAP 2.5° × 2.5° pentad product (Xie and Arkin 1997) because of its longer temporal coverage (1979–present). The monthly and annual climatologies for Singapore and the associated precipitation trends are derived from daily data averaged from 28 stations distributed across the island over the December 1980–November 2014 study period. The ERAI 0000 UTC 850-hPa vector wind field is composited at the native 0.75° × 0.75° spatial resolution for all corresponding cluster days to depict the mean low-level synoptic circulation of each defined regime.

c. Trend detection and estimates

The presence of a monotonic upward or downward trend in both rainfall and regime frequency is analyzed using the nonparametric Mann–Kendall (MK) test, as implemented by the trend_manken function in NCL (NCAR Command Language; http://dx.doi.org/10.5065/D6WD3XH5). This is done by first computing the statistic S, given by

 
formula

where

 
formula

Note that S gives the number of positive differences minus the number of negative differences. Thus, a large positive (negative) value for S means that observations at a later time tend to be larger (smaller) than observations at earlier time. For sample sizes greater than 10, the test statistic is then computed, given by

 
formula

with

 
formula

where VAR(S) is the variance of S, g is the number of tied groups, and is the number of observations in the pth group. A positive (negative) ZMK value indicates an upward (downward) trend. At the level of significance, the null hypothesis of no trend is rejected in favor of the alternate hypothesis of a significant upward (downward) trend if ZMK is positive (negative) and is greater (lesser) than from a standard normal distribution. A statistically significant monotonic trend is detected if the associated probability of ZMK is .

The magnitude or slope of the detected trend is estimated using the nonparametric Theil–Sen slope estimator computed by the trend_manken function. It is given by the median of all pairwise slopes calculated between elements in the time series. We have used the Theil–Sen slope estimator as an estimate of the linear trend because it is more robust and insensitive to outliers compared to the simple least squares regression method (Hess et al. 2001).

3. Characterization and annual distribution of the weather regimes

The climatological daily distribution of the weather states from December to November (Fig. 6) shows distinct, preferred occurrence periods for each regime, with frequency peaks that are relatively well separated from one regime to another in the annual cycle. To further characterize each weather state, the average profiles of total wind vector (m s−1), relative humidity, air temperature anomalies, and equivalent potential temperature for each regime are shown in Fig. 7. The term is a measure of both the temperature and moisture content in the air and therefore an indicator of potential (convective) instability in each weather state. Each regime’s contribution to Singapore rainfall is shown by the local spatial patterns of mean precipitation and 850-hPa wind in Fig. 8. A summary of the mean characteristics for each of the eight weather regimes is provided in Table 1, including their classification acronyms for Singapore. The designated acronyms are based on the mean 850-hPa wind and moisture profiles over Singapore and are therefore only valid when referring to the regimes locally; the regime numbers are additionally used when describing the regional-scale patterns in section 4.

Fig. 6.

Mean annual cycle of the daily relative frequency of the eight weather regimes influencing Singapore rainfall, smoothed by a 31-day centered running average, from 1 Dec to 30 Nov.

Fig. 6.

Mean annual cycle of the daily relative frequency of the eight weather regimes influencing Singapore rainfall, smoothed by a 31-day centered running average, from 1 Dec to 30 Nov.

Fig. 7.

Mean profiles of (a) vector wind, (b) relative humidity, (c) air temperature anomaly, and (d) equivalent potential temperature of the eight regimes over Singapore. The climatological relative humidity and profiles are denoted by the gray short-dashed line.

Fig. 7.

Mean profiles of (a) vector wind, (b) relative humidity, (c) air temperature anomaly, and (d) equivalent potential temperature of the eight regimes over Singapore. The climatological relative humidity and profiles are denoted by the gray short-dashed line.

Fig. 8.

Mean daily precipitation (color) and 850-hPa wind (vectors) over the Singapore region in each of the eight weather regimes. The total number of regime days in the data and the mean regime rainfall based on TMPA 3B42 averaged over the depicted region are shown. Panels are arranged down the columns according to when the main peak of occurrence for each regime is seen in Fig. 6.

Fig. 8.

Mean daily precipitation (color) and 850-hPa wind (vectors) over the Singapore region in each of the eight weather regimes. The total number of regime days in the data and the mean regime rainfall based on TMPA 3B42 averaged over the depicted region are shown. Panels are arranged down the columns according to when the main peak of occurrence for each regime is seen in Fig. 6.

Table 1.

Summary of the mean characteristics for each weather regime. The classification acronyms in the second column are based on the mean 850-hPa wind and moisture profile over Singapore. The values in the square brackets denote the standard deviation.

Summary of the mean characteristics for each weather regime. The classification acronyms in the second column are based on the mean 850-hPa wind and moisture profile over Singapore. The values in the square brackets denote the standard deviation.
Summary of the mean characteristics for each weather regime. The classification acronyms in the second column are based on the mean 850-hPa wind and moisture profile over Singapore. The values in the square brackets denote the standard deviation.

The main discriminators between the regimes over Singapore are low-level wind direction and low-to-midlevel moisture and temperature characteristics (Fig. 7); all regimes generally show easterly winds increasing with height above 500 hPa (Fig. 7a). The regimes are further differentiated by their moisture content over Singapore above 925 hPa. Compared against the climatological RH profile, there are four moister regimes (R4, R6, R1, and R8) and four drier regimes (R7, R3, R2, and R5) (Fig. 7b).

Regimes R3, R4, and R5 depict low-level north-northeasterlies between 1000 and 850 hPa, the prevailing wind direction during the local NE monsoon season. Given that R4 is a very wet regime, it is referred to as the Moist North East (MNE) regime. In contrast, R3 and R5 are both dry and are therefore termed the Dry Northerly (DN) and the Dry North East (DNE) regimes, respectively. Note the wind turns northwesterly at 700–600 hPa in the DN (R3) regime but becomes easterly in the MNE (R4) and DNE (R5) regimes. Interestingly, MNE (R4) is the moistest regime up to 600 hPa while DNE (R5) is the driest (Fig. 7b) despite sharing very similar flow profiles (Fig. 7a) and positive temperature (warm) anomalies above 500 hPa (Fig. 7c). Both the MNE (R4) and DNE (R5) regimes show negative temperature (cool) anomalies below 600 and 850 hPa, respectively, with MNE being cooler than DNE near the surface at 1000 hPa due to cooling from rain evaporation. The wind direction and moisture characteristics are consistent with a very wet and a very dry phase of the local northeast monsoon.

Regimes R2, R7, and R8 exhibit southeasterly-to-southwesterly, southeasterly-to-southerly, and southwesterly flow from 1000 to 850 hPa, respectively. These flow characteristics are typical of the local SW monsoon period. R8 is a fairly wet regime for Singapore while R2 and R7 are both dry regimes. R8 is therefore referred to as the Moist South West (MSW) regime and R2 and R7 are referred to as the Dry South West (DSW) and the Dry Southerly (DS) regimes, respectively. Regime R1 has the most quiescent flow regime, with very little wind up to 600 hPa, while a deep westerly flow dominates in R6 (Fig. 7a). We refer to R1 as the Moist Quiescent (MQ) regime and R6 as the Moist Westerly (MW) regime.

Based on average temperature anomalies below 700 hPa, the MQ (R1), DSW (R2), DS (R7), and MW (R6) regimes are considered as warm weather states for Singapore; the warmest overall is the MQ (R1) regime with positive anomalies throughout the entire depth of the troposphere. DSW is the second warmest regime (but warmest at 1000 hPa) and has a temperature anomaly profile that is the reverse of MNE and DNE; it is warmer below 600 hPa and cooler above, compared to climatology. The cool regimes are DN (R3; coolest below 850 hPa), followed by MNE (R4), DNE (R5) and MSW (R8; only marginally cooler than climatology). A simple matrix summarizing each regime based on its temperature and moisture characteristics below 700 hPa is shown in Table 2. Unsurprisingly, the very warm and moist MQ (R1) regime exhibits the most prime conditions for deep convection relative to climatology, followed by MNE (R4) and MW (R6), which are both deep moist but relatively cooler regimes (Fig. 7d). The much cooler DN (R3) and much drier DNE (R5) are the least favorable to deep convection, relative to climatology. The DSW (R2) regime, while being very warm near the surface, also has low potential (convective) instability because of the very dry conditions, especially above 850 hPa, and likewise for the warm and dry DS (R7) regime, which features slightly less favorable conditions for convection than climatology. The convective potential in MSW (R8) is closest to climatology with its cool but moist lower-to-mid tropospheric environment.

Table 2.

Regime characterization over Singapore relative to the climatological temperature and relative humidity profile averaged between 700 and 1000 hPa.

Regime characterization over Singapore relative to the climatological temperature and relative humidity profile averaged between 700 and 1000 hPa.
Regime characterization over Singapore relative to the climatological temperature and relative humidity profile averaged between 700 and 1000 hPa.

The very moist and cool MNE (R4) regime accounts for 13.1% of all days, peaks in December, and is by far the wettest regime for Singapore. The very high daily rainfall concentrations over the eastern coast and southern tip of the Malay Peninsula identify MNE (R4) as the very wet active phase of the local NE monsoon (Fig. 8a). Cold surges are a feature of the active phase of the NE monsoon in December and January as documented by Lim et al. (2017) and are likely the main contributor to the MNE (R4) regime rainfall.

The MNE peak is then followed in succession by the peaks of DN (R3) in mid-January to mid-February and DNE (R5) in early March (see Fig. 6). DN (R3) exhibits a stronger and more northerly flow than MNE (R4), but is much drier (Fig. 8c). DNE (R5) is the driest NE monsoon regime with daily rainfall below 3 mm day−1 over much of the local domain (Fig. 8e). The dominant appearance of the dry DN regime (R3) and the increasing frequency of the very dry DNE (R5) in February, concomitant with a reduced occurrence of MNE (R4), explain the low climatological rainfall total in February. A month later, the occurrence of MNE’s secondary peak and the appearance of MQ (R1) and MW (R6)—both wet regimes—act to offset the rainfall deficit associated with DNE (R5) and contribute to March total rainfall recovering to January levels (see Fig. 2).

The MQ regime (R1) peaks in May, reaching the highest relative frequency of any weather regime all year round. Its dominance over a 2-month period from April, the weak variable winds over a deep warm layer, and a precipitation pattern that shows more rainfall over land than sea are all hallmarks of the intermonsoon period over Singapore (Fig. 8g). May is also when we begin to see the appearance of the DSW (R2) and DS (R7) regimes, key contributors to the SW monsoon season. DSW is the second warmest and driest regime (based on gauge data) for Singapore. It peaks by mid-June and is the dominant regime for that month, resulting in June having the second lowest monthly total. DSW (R2) is the very dry phase of the local SW monsoon (Fig. 8b) and causes the local boreal summer season to begin with a warm and dry period. Later on, the increasing appearance of MSW (R8) marks the onset of the wet phase for the local SW monsoon over Singapore (Fig. 8d). The wet phase is then maintained through September by the intermittent dominance of both the wet MSW (R8) and the drier DS (R7) regime (Fig. 8f), while the frequency of DSW (R2) diminishes quickly. September also marks the appearance of MW (R6), characterized by deep westerlies, a fairly warm moist environment, and a convectively unstable atmosphere. The MW (R6) regime is the second wettest regime for Singapore and the surrounding local domain (Fig. 8h). The frequency of this regime increases rapidly in October and typically peaks in mid-November as MSW (R8) and DS (R7) occurrences wane. MW (R6) corresponds to the build-up phase of the local NE monsoon and, together with MNE (R4), makes November receive the second highest total rainfall for Singapore.

4. Regional-scale circulation and rainfall patterns over the Maritime Continent

A better appreciation of the role played by the regional-scale structure in terms of the ITCZ position and its interaction with the complex topography of the MC is gained by looking at the mean synoptic-scale rainfall and circulation patterns for each locally identified regime (Fig. 9). As in Fig. 8, these are arranged according to when the main frequency peak of each regime appears (cf. Fig. 6).

Fig. 9.

As in Fig. 8, but for the whole Maritime Continent region.

Fig. 9.

As in Fig. 8, but for the whole Maritime Continent region.

The regimes influencing Singapore’s weather depict very distinct spatial rainfall distributions and notable regional-scale circulation features. The regional-scale patterns can be interpreted as specific phases in the seasonal cross-equatorial migration of the ITCZ over the MC, depicted by the progression of the primary rain belt and related wind flow at 850 hPa.

The MNE (R4) regime is associated with very high daily rainfall amounts over the western MC (Fig. 9a). The highest concentrations are over southeastern Malay Peninsula, the offshore coastal regions along western Sumatra, the northwest coast of Kalimantan, and Java. Strong northeasterly winds south of Vietnam around 8–10 m s−1 are seen over the South China Sea during MNE (R4). This mean characteristic feature would no doubt encompass both full and wind-only cold surge events during the boreal winter monsoon as identified by Lim et al. (2017). In addition, a weak counterclockwise (cyclonic) circulation is seen over the western half of Kalimantan. The counterclockwise turning is likely due to the wind–terrain interaction and conservation of potential vorticity. Interestingly, the circulation is centered in a region known for the formation and high occurrence concentration of the Borneo vortex, an amplified synoptic disturbance feature of the active NE (boreal winter) monsoon examined previously by Chang et al. (2005a), Tangang et al. (2008), and Chen et al. (2015). Its interaction with the prevailing northeasterly 850-hPa wind in the South China Sea would explain the high amount of daily rainfall over the northwest coast of Kalimantan. The location of the heaviest daily rainfall is coincident with the region of enhanced shear vorticity, strong convergence and positive convective index (signifying deep convection) during days when both a cold surge and a Borneo vortex occur together (Chang et al. 2005a; Chen et al. 2015). Given that MNE (R4) occurs most frequently between November and January, it is likely that the MNE (R4) regime would contain most if not all the days in which Borneo vortices developed. Note that very light winds are seen over the Java Sea in MNE (R4), coincident with the position of an equatorial trough/ITCZ being over the region, while easterly winds are characteristic over northern Australia.

During the DN (R3) regime, a strengthening of the wind field over South China Sea and the Java Sea is seen, linked to a strong northerly flow component over Singapore and the Karimata Strait between Sumatra and Kalimantan. This enhanced cross-equatorial flow results in relatively lesser rainfall over the Singapore/Karimata Strait area and corresponds to the dry NE monsoon phase over Singapore (Fig. 9c). The primary rain belt associated with the ITCZ shifts to the Southern Hemisphere and resides over northern Australia, where the highest daily rainfall totals are recorded for that region than in any other regime. The southeastern Philippines also receive their highest daily rainfall in DN (R3) due to stronger winds over the region compared to MNE (R4). Westerly 850-hPa winds over Darwin and a cyclonic circulation over the Top End region are both indicative of the wet and active phase of the austral summer monsoon. The DN (R3) regime over Singapore is therefore likely synonymous with the Deep West wet season regime over northern Australia, as identified by Pope et al. (2009).

In the DNE (R5) regime, the regional-scale flow structure largely resembles that of MNE (R4) but exhibits weakened winds over the eastern Indian Ocean and Indonesia. The regime shows suppressed daily rainfall totals over many land and seas areas relative to MNE (R4) with rainfall largely concentrated over southern Sumatra, western and central Kalimantan, and Java (Fig. 9e). The Malay Peninsula and the South China Sea/Karimata Strait region exhibit very low daily rainfall (≤3 mm day−1). In contrast, the Darwin region and Gulf of Carpentaria over northern Australia both experience enhanced daily rainfall during DNE (R5) compared to MNE (R4).

The MQ (R1) regime represents the transitional (intermonsoon) period between boreal winter and boreal summer monsoon regimes and is dominant between April and June. As in DNE (R5), rainfall is mainly concentrated over land and the coastal areas of Sumatra, the Malay Peninsula, Kalimantan, Sulawesi, and New Guinea, and there is increased daily precipitation over the eastern Indian Ocean due to the prevalence of a cyclonic circulation south of the equator (Fig. 9g). Light and variable winds feature over the Maritime Continent while strong easterly–southeasterly winds are prevalent over latitudes ≥10°S.

Regimes DSW (R2), MSW (R8), and DS (R7) have very similar regional-scale circulation structures (Figs. 9b,d,f), but DSW (R2) shows the strongest zonal flow across South China Sea and Indochina at latitudes ≥6°N that progressively weaken in MSW (R8) and more so in DS (R7), respectively. The main ITCZ rain belt is located at its northernmost extent during DSW (R2) and very dry conditions prevail over the Singapore region (i.e., a very dry SW monsoon phase). The heaviest daily precipitation totals are observed off the coasts of Myanmar and western Philippines (Fig. 9b). During the MSW (R8) regime, zonal westerly winds across Indochina slacken while the westerly winds over South China Sea between 0° and 10°N remain quite strong. The monsoon is still very active over the Northern Hemisphere, but rain activity over the South China Sea is closer to the equator with slightly weaker westerlies compared to DSW (R2), leading to a wet rainfall signature over Singapore (Fig. 9d). In contrast, the monsoonal flow across Indochina and South China Sea is much weaker in the DS (R7) regime relative to MSW (R8) or DSW (R2), and as a result moderate daily rainfall is experienced along the Myanmar coast, over Indochina, along the western coast of Philippines, and over the South China Sea/Singapore region (Fig. 9f). Daily rainfall is also relatively suppressed over the western North Pacific. Meanwhile, rainfall over the eastern Indian Ocean and the western coast of Sumatra is high.

The MW (R6) regime is characterized by a large cyclonic vortex over the South China Sea (SCS) north of Kalimantan, coincident with high daily rainfall there. In addition, strong westerlies from the Indian Ocean sweep across the western and central Maritime Continent between 4°N and 3°S (Fig. 9h). The flow is easterly across the Philippines and over northern Australia. Remarkably, the vortex feature in MW (R6) suggests that the regime captures the climatological spatial distribution of SCS vortices during November (cf. Fig. 3 of Nguyen et al. 2016). This indicates that a SCS vortex is an important regional circulation feature bringing high rainfall to Vietnam (Nguyen et al. 2016) and especially to the northeast coast of the Malay Peninsula due to the convergent flow around the western flank of the vortex. Rainfall amounts there are the highest of anywhere in the Maritime Continent during MW (R6). The overall setup of the regional-scale flow results in a double ITCZ structure over the Maritime Continent that is very symmetric over the equator: the northern limb is around 5°–6°N while the southern limb is around 6°S. On average, the SCS vortex is most likely formed by the interaction between the orographically deflected near-equatorial westerlies over Kalimantan and the southern portion of the easterly flow over Vietnam, but may also include tropical storms that cross southern Philippines, propagating westward and interacting with the NE monsoonal flow.

The moisture advected by the deep westerly flow leads to a more potentially unstable environment over the equatorial region relative to climatology. In addition, the predominant occurrence of the MW (R6) regime explains why it is wet over Singapore and much of Sumatra and Borneo during October and November, prior to the proper arrival of the local NE monsoon. The deep moist westerly flow of MW (R6) also partly contributes to the high daily rainfall over Singapore in April and May by preconditioning the atmosphere for deep convection. The predominance of the MW (R6) weather state in October and November, with its strong deep westerly flow, explains why Sumatra squall activity across Singapore is most frequent in those months followed by April and May (Lo and Orton 2016).

At this stage, a comparison with the weather types (WTs) obtained by Moron et al. (2015) is instructive. When the frequency distribution and the associated spatial maps of their WTs [see Figs. 3 and 4 of Moron et al. (2015)] are compared with those shown in Figs. 6 and 9, we find that five of our weather regimes qualitatively map out and correspond to those obtained by Moron et al. (2015). Specifically, their WT1 corresponds to MSW (R8), WT3 is similar to MNE (R4), WT4 resembles DN (R3), WT5 is akin to DNE (R5), and WT6 looks like MQ (R1). The similitude of the regional spatial patterns found between the two studies suggests that the physical regime structures are robust and stable, independent of the different reanalysis data (NCEP2 vs ERA-Interim) and the approach taken (clustering of horizontal wind field versus soundings).

5. Regime and rainfall trends

Singapore has experienced a statistically significant upward trend in annual December–November rainfall since 1981. While few are statistically significant, trends are also evident in most calendar months (Fig. 3). In this section, we investigate how these annual rainfall trends may be related to either changes in regime frequency or changes to within-regime rainfall itself.

The annual (December–November) frequencies of each regime (Fig. 10) show statistically significant upward MW (R6) and downward DN (R3) trends at the 5% level, and also an upward trend in DSW (R2) significant at the 10% level. All other regimes only have weak trends: upward for MQ (R1) and MNE (R4) or downward for DNE (R5), DS (R7), and MSW (R8). Overall, the relative frequency of dry regimes for Singapore has generally decreased [led by DN (R3)], while the relative occurrence of wet regimes has increased [dominated by MW (R6)]. The shift toward more frequent wet regimes is only partly offset by an increased occurrence in DSW (R2) (the driest regime according to gauge data).

Fig. 10.

Annual (December–November) regime frequency of occurrence from 1981 to 2014 (in days yr−1) with estimated linear slopes (red lines) shown. A significant monotonic trend at the 5% level is detected if the associated probability P from the Mann–Kendall test is P ≥ 0.95.

Fig. 10.

Annual (December–November) regime frequency of occurrence from 1981 to 2014 (in days yr−1) with estimated linear slopes (red lines) shown. A significant monotonic trend at the 5% level is detected if the associated probability P from the Mann–Kendall test is P ≥ 0.95.

Both DSW (R2) and DNE (R5) show very small upward trends in within-regime rainfall for Singapore. This is in contrast to MW (R6), which shows a statistically significant upward trend (Fig. 11). MSW (R8) also depicts an increasing rainfall trend that is statistically significant at the 10% level. Over the same period, DN (R3) has become an even drier regime for Singapore, while MQ (R1), MNE (R4), and DS (R7) have become intrinsically wetter, although their respective trends are not statistically significant. The positive frequency trend in DSW (R2) and the small to moderate enhancements in intrinsic regime rainfall from DSW (R2), DS (R7), and MSW (R8) combine to indicate that, over the study period, the local SW monsoon rainfall activity over Singapore has slightly strengthened. Overall, the observed rainfall increase within the wetter weather regimes—MQ (R1), MNE (R4), MW (R6), and MSW (R8)—is consistent with the expected rise in atmospheric moisture due to ongoing global warming.

Fig. 11.

Annual mean rainfall over Singapore in each regime (expressed in mm day−1). The average daily regime rainfall over the 1981–2014 period is also indicated.

Fig. 11.

Annual mean rainfall over Singapore in each regime (expressed in mm day−1). The average daily regime rainfall over the 1981–2014 period is also indicated.

Following Catto et al. (2012), who studied wet season precipitation trends over Darwin, Australia, we have decomposed the total annual trend using the following equation:

 
formula

where i denotes the regime number, is the mean frequency of occurrence (in days yr−1) for regime i, is the mean precipitation amount within regime i (mm day−1), and and represent the linear trends in the two measures. The term reflects the contribution from changes in within-regime precipitation while the term represents the contribution from changes in regime frequency. The third term denotes the residual contribution from the two linear trends; this term is very small and therefore can be ignored. Note the trend decomposition applied here follows strictly the method of Catto et al. (2012) and is done with least squares linear regression.

Figure 12 summarizes the relative extent to which changes in occurrence frequency and within-regime precipitation from each regime contribute to the overall annual rainfall trend (in mm yr−1). The net rainfall increase in MQ (R1) and MW (R6) is largely driven by those regimes happening more often, while the reverse is true for DN (R3) and DNE (R5), both dry regimes. Meanwhile, the positive rainfall trend in MNE (R4) is due to enhanced within-regime rainfall and more regime occurrences in almost equal measures. This indicates that there is a shift toward more frequent wetter days (and fewer dry days) during the local NE monsoon and intermonsoon seasons. For the SW monsoon period, the trend in DSW (R2) is equally driven by both increased regime frequency and within-regime precipitation. In contrast, the increases in within-regime rainfall in DS (R7) and MSW (R8) are offset by the reduced occurrence of those regimes, such that the net trend is near zero in those regimes.

Fig. 12.

Contributions from each regime to the overall annual rainfall trend (mm yr−1) at Singapore. For each regime, the component related to the changes in regime frequency (; orange bars) and the component associated with changes in within-regime precipitation (; blue bars) are shown. The absolute magnitude of changes from the two trend components are shown by the bold text at the bottom right-hand corner, respectively.

Fig. 12.

Contributions from each regime to the overall annual rainfall trend (mm yr−1) at Singapore. For each regime, the component related to the changes in regime frequency (; orange bars) and the component associated with changes in within-regime precipitation (; blue bars) are shown. The absolute magnitude of changes from the two trend components are shown by the bold text at the bottom right-hand corner, respectively.

Notably, the breakdown shows that slightly more than two-thirds of the observed trend is driven by the net rainfall increase due to changes in within-regime precipitation (shown by the blue bar of the Rsum value), with all regimes depicting enhanced within-regime rainfall except for DN (R3) and DNE (R5). However, this apparent result neglects a much larger contribution resulting from changes in regime frequency (orange bars); these actually dominate when the absolute magnitude of the individual terms are computed (i.e., ). Note that gives a negligible contribution across all the regimes.

Net and absolute contributions from the two main trend components are shown for the MC region in Fig. 13. Here the CMAP product has been utilized to calculate the linear annual trends in precipitation. The net rainfall trend due to changes in regime frequency is small (mostly between 0 to 4 mm yr−1) and fairly homogeneous across the region (Fig. 13a). Interestingly, apart from the eastern (western) coast of southern Philippines (Kalimatan), net increases are largely seen in the Northern Hemisphere, while net decreases are evident in the Southern Hemisphere, except for the south Indian Ocean region roughly between 90° and 105°E. In contrast, the net contribution due to changes in within-regime rainfall depicts more spatial heterogeneity (Fig. 13c) and closely resembles the pattern seen in Fig. 4. Overall, changes in within-regime rainfall constitute a large proportion of the observed rainfall trend (≥60%) across many MC locations when the ratio between the two trend components is calculated (Fig. 13e). There are intriguing exceptions, however, where the observed rainfall trend is mainly driven by changes in regime frequency (e.g., the Bay of Bengal).

Fig. 13.

(a) Net rainfall change due to changes in regime frequency. (b) Absolute change in rainfall due to changes in regime frequency. (c) Net rainfall change due to changes in within-regime precipitation. (d) Absolute change in rainfall due to changes in within-regime precipitation. (e) Ratio between (a) and (c); absolute values are used to comparatively highlight the size of the net contribution from each trend component. Values above (below) 1 indicate that the net rainfall change due to regime frequency changes is greater (less) than that from changes in within-regime rainfall. (f) Ratio between (b) and (d); values above (below) 1 denote that circulation (thermodynamic) changes amount to a larger contribution in absolute terms.

Fig. 13.

(a) Net rainfall change due to changes in regime frequency. (b) Absolute change in rainfall due to changes in regime frequency. (c) Net rainfall change due to changes in within-regime precipitation. (d) Absolute change in rainfall due to changes in within-regime precipitation. (e) Ratio between (a) and (c); absolute values are used to comparatively highlight the size of the net contribution from each trend component. Values above (below) 1 indicate that the net rainfall change due to regime frequency changes is greater (less) than that from changes in within-regime rainfall. (f) Ratio between (b) and (d); values above (below) 1 denote that circulation (thermodynamic) changes amount to a larger contribution in absolute terms.

Similar to Singapore, the gross contribution from regime frequency changes outweighs the contribution due to changes in within-regime precipitation when the absolute magnitude of each regime term is considered (Figs. 13b,d). This is expressed by the ratio in Fig. 13f. Notably, there are regions such as the Philippines, Sulawesi, New Guinea, and the coastal parts of Vietnam, Thailand, and Myanmar where changes in within-regime rainfall dominate.

The above analysis implies that, in absolute terms, changes in the regional-scale circulation are more important than altered thermodynamic conditions over large swaths of the MC. This includes the region around Darwin and is consistent with the results reported by Catto et al. (2012) for the austral summer monsoon over a longer period. Interestingly, the rest of continental northern Australia show overall rainfall trends that are largely driven by thermodynamic changes, as do many locations south of 12°S in the southern Indian Ocean.

6. Discussion and conclusions

The Maritime Continent is a region that experiences high rainfall variability on many time scales. In many areas, the weather experienced daily is characterized by small-scale convective systems that are often embedded within a larger envelope of convective activity. It is also possible for locations in this region to identify seasons based on the main circulation patterns related to the interhemispheric passage of the intertropical convergence zone (ITCZ) that drives the Asian–Australian monsoon system. For Singapore, this corresponds to the northeast monsoon during boreal winter, the southwest monsoon during boreal summer, and the intermonsoon seasons. In this study, a methodology was developed to classify daily weather maps based on a clustering of eight clearly separated regional-scale weather regimes that influence Singapore rainfall annually. The weather regimes are derived by k-means clustering of ERA-Interim (ERAI) data for the period December 1980 to November 2014, using sounding profiles of winds, temperature and specific humidity extracted over the grid point corresponding to Singapore. Vertical soundings from ERAI were preferred to actual radiosonde measurements made over Singapore due to the inhomogeneities existing in the observations. In addition, the use of reanalysis data, which come from a more physically consistent dataset, means that such a methodology can be replicated elsewhere for a location where sounding observations are not readily available.

These regional-scale regimes are meaningful in terms of the type of weather experienced locally at different times of the year, either wet or dry and cool or warm. Each regime displays a clear and distinct frequency distribution in terms of its respective annual cycle of occurrences. Indeed, the eight weather regimes identified here correspond to the seasonal migration of the ITCZ across the equator and provide a unified picture of the mean annual cycle of rainfall for Singapore and the Maritime Continent. In that regard, this study extends previous work by Moron et al. (2015) by identifying weather regimes all year round. Five of our eight regimes qualitatively map to those defined in their study. The remarkable similarities of the corresponding regional spatial patterns of mean 850-hPa wind and rainfall/OLR between our two studies suggest that the physical regime structures found here are robust and stable, despite being defined from a single grid point (Singapore) in the Maritime Continent. The other three regimes complement their earlier findings, which did not cover the full annual cycle.

The weather regimes are used to explain observed annual trends in Singapore rainfall (from rain gauges) and across the Maritime Continent (from the CMAP satellite rainfall product). Singapore has experienced a statistically significant increase in annual rainfall since 1980 to 2014 as measured by gauges, as do areas around Singapore estimated by the CMAP product. Positive trends for Singapore are also apparent in 9 months of the year according to the rain gauge network, while all 12 months show positive trends in the satellite product for rainfall averaged around Singapore.

In an annual sense, the observed rainfall trend at Singapore and many other Maritime Continent locations is primarily driven by changes in within-regime precipitation. For Singapore, all weather regimes show a consistent increase in intrinsic rainfall amount apart from DN (R3), which showed a reduction. This supports the thermodynamic Clausius–Clapeyron (CC) argument that with increased temperatures the atmosphere is capable of holding more moisture, thus potentially delivering more precipitation when it rains. However, this net result conceals the fact that for most weather regimes there is a larger absolute contribution stemming from changes in the regime frequency (i.e., circulation changes) than that solely due to changes in intrinsic precipitation amount within the regime itself. This factor is also highlighted across many other locations in the Maritime Continent, showing the dominance of regime frequency changes in an absolute sense. Nonetheless, contributions to the overall rainfall trend due to changes in regime frequency tend to offset one another in reality, reducing their net effect.

Another very important finding is that the only two weather regimes that contribute negatively to the overall local rainfall trend are DNE (R5) and DN (R3), which are the second and third driest regimes for Singapore, respectively. These regimes occur largely during the northeast monsoon season between December–April. Concurrently, the three wettest regimes for Singapore—MNE (R4), MQ (R1), and MW (R6)—give the largest positive trends. The trends in MQ (R1) and MW (R6) are largely driven by increased regime frequency, while the trend in MNE (R4) is driven by both enhanced within-regime precipitation and more regime occurrences in almost equal parts. The fact that the wet regimes [MNE (R4) and MW (R6)] are getting wetter and occurring more often while the dry regimes [DN (R3) and DNE (R5)] are getting drier and less frequent means that rainfall variability over Singapore has been enhanced during the boreal winter monsoon season. This result, based on objectively identified weather regimes, can be interpreted as an illustration of the “wet get wetter and dry get drier” argument about the possible impact of global warming on the hydrological cycle.

Nevertheless, climate change model simulations are inconclusive regarding the possible future projections of annual rainfall totals for Singapore. The Second Singapore Climate Change Study shows a considerable spread between regional climate models [see Fig. 4.6 in CCRS (2015)] with possibilities of an increase or decrease in mean annual rainfall by the end of the twenty-first century. This suggests that circulation changes—leading to a redistribution of moisture—could play a pivotal role in addition to thermodynamic changes related to the CC argument with regards to rainfall over the Maritime Continent. In other words, the range of model responses is likely to be driven in part by different circulation changes among the climate models and should not be neglected when considering future precipitation trends.

Thus, an interesting extension of this study would be to apply the concept weather regimes as a tool to evaluate climate models. Assessing the ability of models to simulate observed regimes in terms of their relative frequency and regional structure would first help provide confidence in the models’ abilities. This, in turn, would allow rainfall projections to be assessed in terms of either a change of rainfall within a regime (i.e., thermodynamic changes) and/or changes in regime frequency (i.e., circulation changes). This methodology offers an interesting prospect to further understand rainfall projections within the region and enhance the confidence in model results.

Acknowledgments

The authors thank the editor and the two reviewers for their comments and suggestions that helped improve the quality of the manuscript. Data processing, analysis, and visualization were collectively done with the suite of Climate Data Operators (version 1.9.1) from the Max-Planck-Institut für Meteorologie and the NCAR Command Language (version 6.4.0; UCAR/NCAR/CISL/VETS; https://doi.org/10.5065/D6WD3XH5).

REFERENCES

REFERENCES
Birch
,
C. E.
,
S.
Webster
,
S. C.
Peatman
,
D. J.
Parker
,
A. J.
Matthews
,
Y.
Li
, and
M. E. E.
Hassim
,
2016
:
Scale interactions between the MJO and the western Maritime Continent
.
J. Climate
,
29
,
2471
2492
, https://doi.org/10.1175/JCLI-D-15-0557.1.
Catto
,
J. L.
,
C.
Jakob
, and
N.
Nicholls
,
2012
:
The influence of changes in synoptic regimes on north Australian wet season rainfall trends
.
J. Geophys. Res.
,
117
,
D10102
, https://doi.org/10.1029/2012JD017472.
CCRS
,
2015
: Singapore’s Second National Climate Change Study: Climate Projections to 2100. Tech. Rep., Centre for Climate Research Singapore, 73 pp., http://ccrs.weather.gov.sg/wp-content/uploads/2015/04/Publications-Second-National-Climate-Change-Study-Report-for-Stakeholders.pdf.
Chang
,
C.-P.
,
Z.
Wang
,
J.
Ju
, and
T.
Li
,
2004
:
On the relationship between western Maritime Continent monsoon rainfall and ENSO during northern winter
.
J. Climate
,
17
,
665
672
, https://doi.org/10.1175/1520-0442(2004)017<0665:OTRBWM>2.0.CO;2.
Chang
,
C.-P.
,
P. A.
Harr
, and
H.-J.
Chen
,
2005a
:
Synoptic disturbances over the equatorial South China Sea and western Maritime Continent during boreal winter
.
Mon. Wea. Rev.
,
133
,
489
503
, https://doi.org/10.1175/MWR-2868.1.
Chang
,
C.-P.
,
Z.
Wang
,
J.
McBride
, and
C.-H.
Liu
,
2005b
:
Annual cycle of Southeast Asia–Maritime Continent rainfall and the asymmetric monsoon transition
.
J. Climate
,
18
,
287
301
, https://doi.org/10.1175/JCLI-3257.1.
Chen
,
T. C.
,
J. D.
Tsay
,
J.
Matsumoto
, and
J.
Alpert
,
2015
:
Development and formation mechanism of the Southeast Asian winter heavy rainfall events around the South China Sea. Part I: Formation and propagation of cold surge vortex
.
J. Climate
,
28
,
1417
1443
, https://doi.org/10.1175/JCLI-D-14-00170.1.
Dee
,
D. P.
, and Coauthors
,
2011
:
The ERA-Interim reanalysis: Configuration and performance of the data assimilation system
.
Quart. J. Roy. Meteor. Soc.
,
137
,
553
597
, https://doi.org/10.1002/qj.828.
Hartigan
,
J.
, and
M.
Wong
,
1979
:
Algorithm AS 136: A K-means clustering algorithm
.
Appl. Stat.
,
28
,
100
108
, https://doi.org/10.2307/2346830.
Haylock
,
M.
, and
J.
McBride
,
2001
:
Spatial coherence and predictability of Indonesian wet season rainfall
.
J. Climate
,
14
,
3882
3887
, https://doi.org/10.1175/1520-0442(2001)014<3882:SCAPOI>2.0.CO;2.
Hertig
,
E.
, and
J.
Jacobeit
,
2014
:
Variability of weather regimes in the North Atlantic-European area: Past and future
.
Atmos. Sci. Lett.
,
15
,
314
320
, https://doi.org/10.1002/asl2.505.
Hess
,
A.
,
H.
Iyer
, and
W.
Malm
,
2001
:
Linear trend analysis: A comparison of methods
.
Atmos. Environ.
,
35
,
5211
5222
, https://doi.org/10.1016/S1352-2310(01)00342-9.
Hess
,
P.
, and
H.
Brezowsky
,
1952
: Katalog der Grosswetterlagen Europas (catalog of European large scale weather types). Berichte des Deutschen Wetterdienstes in der US-Zone 33, 39 pp.
Huffman
,
G. J.
, and Coauthors
,
2007
:
The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales
.
J. Hydrometeor.
,
8
,
38
55
, https://doi.org/10.1175/JHM560.1.
Huffman
,
G. J.
,
R. F.
Adler
,
D. T.
Bolvin
, and
E. J.
Nelkin
,
2010
: The TRMM Multi-Satellite Precipitation Analysis (TMPA). Satellite Rainfall Applications for Surface Hydrology, M. Gebremichael and F. Hossain, Eds., Springer, 3–22, https://doi.org/10.1007/978-90-481-2915-7_1.
Jiang
,
N.
,
W.
Qian
, and
J. C.-H.
Leung
,
2016
:
The global monsoon division combining the k-means clustering method and low-level cross-equatorial flow
.
Climate Dyn.
,
47
,
2345
2359
, https://doi.org/10.1007/s00382-015-2967-5.
Kikuchi
,
K.
, and
B.
Wang
,
2008
:
Diurnal precipitation regimes in the global tropics
.
J. Climate
,
21
,
2680
2696
, https://doi.org/10.1175/2007JCLI2051.1.
Lamb
,
H. H.
,
1972
: British Isles weather types and a register of daily sequence of circulation patterns, 1861–1971. Met OfficeGeophysical Memoirs, Vol. 116, H. M. Stationery Office, 85 pp.
Lestari
,
R. K.
,
M.
Watanabe
,
Y.
Imada
,
H.
Shiogama
,
R. D.
Field
,
T.
Takemura
, and
M.
Kimoto
,
2015
:
Increasing potential of biomass burning over Sumatra, Indonesia induced by anthropogenic tropical warming
.
Environ. Res. Lett.
,
9
, 104010, https://doi.org/10.1088/1748-9326/9/10/104010.
Liebmann
,
B.
, and
H.
Hendon
,
1990
:
Synoptic-scale disturbances near the equator
.
J. Atmos. Sci.
,
47
,
1463
1479
, https://doi.org/10.1175/1520-0469(1990)047<1463:SSDNTE>2.0.CO;2.
Lim
,
S. Y.
,
C.
Marzin
,
P.
Xavier
,
C.-P.
Chang
, and
B.
Timbal
,
2017
:
Impacts of boreal winter monsoon cold surges and the interaction with MJO on Southeast Asia rainfall
.
J. Climate
,
30
,
4267
4281
, https://doi.org/10.1175/JCLI-D-16-0546.1.
Lo
,
J. C. F.
, and
T.
Orton
,
2016
:
The general features of tropical Sumatra squalls
.
Weather
,
71
,
175
178
, https://doi.org/10.1002/wea.2748.
McBride
,
J. L.
,
M. R.
Haylock
, and
N.
Nicholls
,
2003
:
Relationships between the Maritime Continent heat source and the El Niño–Southern Oscillation phenomenon
.
J. Climate
,
16
,
2905
2914
, https://doi.org/10.1175/1520-0442(2003)016<2905:RBTMCH>2.0.CO;2.
Michelangeli
,
R. A.
,
R.
Vautard
, and
B.
Legras
,
1995
:
Weather regimes: Recurrence and quasi-stationarity
.
J. Atmos. Sci.
,
52
,
1237
1256
, https://doi.org/10.1175/1520-0469(1995)052<1237:WRRAQS>2.0.CO;2.
Moron
,
V.
,
A. W.
Robertson
,
M. N.
Ward
, and
O.
Ndiaye
,
2008
:
Weather types and rainfall over Senegal. Part I: Observational analysis
.
J. Climate
,
21
,
266
287
, https://doi.org/10.1175/2007JCLI1601.1.
Moron
,
V.
,
A. W.
Robertson
, and
J.-H.
Qian
,
2010
:
Local versus regional-scale characteristics of monsoon onset and post-onset rainfall over Indonesia
.
Climate Dyn.
,
34
,
281
299
, https://doi.org/10.1007/s00382-009-0547-2.
Moron
,
V.
,
A. W.
Robertson
,
J.-H.
Qian
, and
M.
Ghil
,
2015
:
Weather types across the Maritime Continent: From the diurnal cycle to interannual variations
.
Front. Environ. Sci.
,
2
,
65
, https://doi.org/10.3389/fenvs.2014.00065.
Nguyen
,
D.-Q.
,
J.
Renwick
, and
J.
McGregor
,
2016
:
On the presence of tropical vortices over the Southeast Asian sea–Maritime Continent region
.
J. Climate
,
29
,
4793
4800
, https://doi.org/10.1175/JCLI-D-14-00468.1.
Oh
,
J.-H.
,
K.-Y.
Kim
, and
G.-H.
Lim
,
2012
:
Impact of MJO on the diurnal cycle of rainfall over the western Maritime Continent in the austral summer
.
Climate Dyn.
,
38
,
1167
1180
, https://doi.org/10.1007/s00382-011-1237-4.
Peatman
,
S. C.
,
A. J.
Matthews
, and
D. P.
Stevens
,
2014
:
Propagation of the Madden-Julian Oscillation through the Maritime Continent and scale interaction with the diurnal cycle of precipitation
.
Quart. J. Roy. Meteor. Soc.
,
140
,
814
825
, https://doi.org/10.1002/qj.2161.
Pohl
,
B.
,
P.
Camberlin
, and
P.
Roucou
,
2005
:
Typology of pentad circulation anomalies over the eastern Africa–western Indian Ocean region, and their relationship with rainfall
.
Climate Res.
,
29
,
111
127
, https://doi.org/10.3354/cr029111.
Pope
,
M.
,
C.
Jakob
, and
M. J.
Reeder
,
2009
:
Regimes of the north Australian wet season
.
J. Climate
,
22
,
6699
6715
, https://doi.org/10.1175/2009JCLI3057.1.
Qian
,
J.-H.
,
2008
:
Why precipitation is mostly concentrated over islands in the Maritime Continent
.
J. Atmos. Sci.
,
65
,
1428
1441
, https://doi.org/10.1175/2007JAS2422.1.
Qian
,
J.-H.
,
A. W.
Robertson
, and
V.
Moron
,
2010
:
Interactions among ENSO, the monsoon, and diurnal cycle in rainfall variability over Java, Indonesia
.
J. Atmos. Sci.
,
67
,
3509
3524
, https://doi.org/10.1175/2010JAS3348.1.
Qian
,
J.-H.
,
A. W.
Robertson
, and
V.
Moron
,
2013
:
Diurnal cycle in different weather regimes and rainfall variability over Borneo associated with ENSO
.
J. Climate
,
26
,
1772
1790
, https://doi.org/10.1175/JCLI-D-12-00178.1.
Rauniyar
,
S. P.
, and
K. J. E.
Walsh
,
2011
:
Scale interaction of the diurnal cycle of rainfall over the Maritime Continent and Australia: Influence of the MJO
.
J. Climate
,
24
,
325
348
, https://doi.org/10.1175/2010JCLI3673.1.
Rauniyar
,
S. P.
,
A.
Protat
, and
H.
Kanamori
,
2017
:
Uncertainties in TRMM-Era multisatellite-based tropical rainfall estimates over the Maritime Continent
.
Earth Space Sci.
,
4
,
275
302
, https://doi.org/10.1002/2017EA000279.
Robertson
,
A. W.
, and
M.
Ghil
,
1999
:
Large-scale weather regimes and local climate over the western United States
.
J. Climate
,
12
,
1796
1813
, https://doi.org/10.1175/1520-0442(1999)012<1796:LSWRAL>2.0.CO;2.
Robertson
,
A. W.
,
V.
Moron
,
J.-H.
Qian
,
C.-P.
Chang
,
F. T.
Tangang
,
E.
Aldrian
,
T.-Y.
Koh
, and
J.
Liew
,
2011
: The Maritime Continent monsoon. The Global Monsoon System: Research and Forecast, C. P. Chang et al., Eds., World Scientific, 85–98, https://doi.org/10.1142/9789814343411_0006.
Tangang
,
F. T.
,
L.
Juneng
,
E.
Salimun
,
P. N.
Vinayachandran
,
Y. K.
Seng
,
C. J. C.
Reason
,
S. K.
Behera
, and
T.
Yasunari
,
2008
:
On the roles of the northeast cold surge, the Borneo vortex, the Madden–Julian oscillation, and the Indian Ocean dipole during the extreme 2006/2007 flood in southern peninsular Malaysia
.
Geophys. Res. Lett.
,
35
,
L14S07
, https://doi.org/10.1029/2008GL033429.
Vautard
,
R.
,
1990
:
Multiple weather regimes over the North Atlantic: Analysis of precursors and successors
.
Mon. Wea. Rev.
,
118
,
2056
2081
, https://doi.org/10.1175/1520-0493(1990)118<2056:MWROTN>2.0.CO;2.
Vautard
,
R.
, and
B.
Legras
,
1988
:
On the source of midlatitude low-frequency variability. Part II: Nonlinear equilibration of weather regimes
.
J. Atmos. Sci.
,
45
,
2845
2867
, https://doi.org/10.1175/1520-0469(1988)045<2845:OTSOML>2.0.CO;2.
Vincent
,
C. L.
, and
T. P.
Lane
,
2016
:
Evolution of the diurnal precipitation cycle with the passage of a Madden–Julian oscillation event through the Maritime Continent
.
Mon. Wea. Rev.
,
144
,
1983
2005
, https://doi.org/10.1175/MWR-D-15-0326.1.
Xie
,
P.
, and
P. A.
Arkin
,
1997
:
Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs
.
Bull. Amer. Meteor. Soc.
,
78
,
2539
2558
, https://doi.org/10.1175/1520-0477(1997)078<2539:GPAYMA>2.0.CO;2.
Yamanaka
,
M. D.
,
2016
:
Physical climatology of Indonesian maritime continent: An outline to comprehend observational studies
.
Atmos. Res.
,
178-179
,
231
259
, https://doi.org/10.1016/j.atmosres.2016.03.017.
Yang
,
G.-Y.
, and
J.
Slingo
,
2001
:
The diurnal cycle in the tropics
.
Mon. Wea. Rev.
,
129
,
784
801
, https://doi.org/10.1175/1520-0493(2001)129<0784:TDCITT>2.0.CO;2.
Yiou
,
P.
, and
M.
Nogaj
,
2004
:
Extreme climatic events and weather regimes over the North Atlantic: When and where?
Geophys. Res. Lett.
,
31
,
L07202
, https://doi.org/10.1029/2003GL019119.

Footnotes

For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).