1. Introduction
The Asian summer monsoon is a large system with well-defined regional patterns and seasonal signals. It can bring heavy rainfall from late spring to autumn, depending on the region, and is followed by dry winters. Its dynamics have been reviewed in many papers (e.g., Ramage 1971; Ding 1994; Chang et al. 2005; Wang 2006; Ding 2007). The Asian summer monsoon provides necessary water for agriculture and societal needs but can also bring extreme precipitation, often leading to floods. Furthermore, droughts can be problematic during the dry season. Future projections of precipitation in a warming climate and the associated mechanism have been examined in various studies (Chou and Neelin 2004; Min et al. 2006; Kripalani et al. 2007; Stephens and Ellis 2008; Chou et al. 2009; Wu et al. 2009; Seager et al. 2010; Chou et al. 2012; Kusunoki and Arakawa 2012; Hsu et al. 2012; Ma et al. 2012; Wu and Chou 2012; Lee and Wang 2014; Seo et al. 2013), and the variations in the areas affected by monsoon as well as monsoon dynamics have been subjects of research (Wang and Ding 2006; Inoue and Ueda 2011; Min et al. 2012; Turner and Annamalai 2012; Duan et al. 2013; Hsu et al. 2013; Jones and Carvalho 2013; Seth et al. 2013). One of the major phenomena linked with the global climate change is the evolution of extreme hydrological events. Various studies have indicated global and regional variations of these extreme events in a warming climate, using past observations and numerical projections (Trenberth et al. 2003; Kharin and Zwiers 2005; Meehl et al. 2005; Räisänen 2005; Barnett et al. 2006; Tebaldi et al. 2006; Giorgi et al. 2011; Shiu et al. 2012; Scoccimarro et al. 2013). The Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4) provides a summary of these studies; for example, chapters 10.3.6 (Meehl et al. 2007) and 11.4 (Christensen et al. 2007) detail projections for the Asian region.
In this study, we investigate the regional behavior of extreme rainfall and the change in the distribution of precipitation in the context of global warming, and the possible large-scale atmospheric factors associated with these changes. We analyze monthly signal and spatial patterns of extreme precipitation behavior, during specific phases of the Asian monsoon, as well as changes in the dynamics of the lower atmosphere. Although the findings are mostly relevant to the summer monsoon, we derive several results related to wintertime, thus representing an investigation of the changes in seasonal signals in Asia. We use an ensemble of 30 models from phase 5 of the Coupled Model Intercomparison Project (CMIP5), following the representative concentration pathway 8.5 (RCP8.5). The ensemble method enables identifying the main uncertainties and the most confident results concerning the change in hydrological dynamics.
The data and methodology used are described in section 2. Section 3 presents an analysis of the confidence of the projection, using the chosen model ensemble. Section 4 provides details of the change of extreme events during the Asian summer monsoon and their link with the dynamics. Section 5 presents some discussion and the conclusions.
2. Data and methodology
a. Data
In this study, we used a single member of 30 different models from the CMIP5 multimodel ensemble (listed in Table 1 with their respective resolution). Each model was first forced by historical forcing (observation of aerosols, greenhouse gases, and solar irradiance) until 2005, and then followed the RCP8.5 pathway (Riahi et al. 2007), leading to an increase of 8.5 W m−2 of radiative forcing by the end 2100. We considered two periods for comparison: from 1976 to 2005 (HIST) and from 2071 to 2100 (RCP). We used daily data for the precipitation, specific humidity, winds, and monthly data for the temperature. The results are presented as an average of the 30 years of each period.
30 CMIP5 models used for this study. The resolution is given in grid points (latitude × longitude). Asterisks indicate models used for the analysis in section 4.
Figure 1 displays the mean precipitation during the Asian summer monsoon (May–August), obtained from the daily (1997–2007) Global Precipitation Climatology Project (GPCP) observations (Huffman et al. 2001). Based on the specific dynamics and spatial features associated with the Asian summer monsoon, we separated the entire domain into three subregions, each corresponding to a main phase of the East Asian summer monsoon region (EAR: 22°–45°N, 105°–145°E), India summer monsoon region (IR: 5°–28°N, 70°–105°E), and western North Pacific summer monsoon region (WNPR: 5°–22°N, 105°–160°E). The definition of these regions is similar to that of Ding (2007).
Mean May–August precipitation from GPCP (Huffman et al. 2001) daily data (1997–2007), and illustration of the three main domains used for this study: EAR (22°–45°N, 105°–145°E), IR (5°–28°N, 70°–105°E), and WNPR (5°–22°N, 105°–160°E).
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
b. Methodology
Our analysis mostly focuses on the change in intensity and frequency of precipitation, defined below.
1) Intensity
Intensity is directly computed by separating the precipitation into percentiles, at each grid point. Values are computed by steps of 10 between 1 and 90, and by steps of 1 between 90 and 99. We compute either yearly percentiles (using the entire 30 yr), or monthly percentiles (using only the precipitation from identical months, over the period of 30 yr). The intensity of a percentile X is denoted pctX and is expressed in mm day−1.
2) Frequency
As a threshold, for each grid point, we use the HIST intensity value of a given yearly percentile, averaged over a 10° × 10° box around this point. This threshold is independent from the months. Also note that HIST and RCP use the same threshold value for a given percentile. For each grid point, each day with precipitation higher than this threshold is counted as an event (for the given percentile). Various methods are feasible, but this unique threshold enables identifying the periods of the highest precipitation (the summer monsoon). Because we use the same threshold for HIST and RCP, the frequency for each period corresponds to the same type of event. In other words, we are observing the change of frequency of events defined as extremes in HIST. Finally, we compute the monthly frequency by averaging the count of events from the same months over the period of 30 years. The frequency of a percentile X is denoted fqpctX and is expressed as a number of days.
We insist on the specific computation for intensity and frequency. Intensity is computed directly for each month and grid point, and for the two periods (HIST and RCP). Frequency uses the same threshold for HIST and RCP (independent from the month) but is computed for each month and grid point, and for the two periods (HIST and RCP). Therefore, both pctX and fqpctX values exhibited time and space variations, but fqpctX is based on the HIST threshold.
We also define the change as the difference between RCP and HIST, RCP − HIST, and the relative change as the difference normalized by HIST, 
3. Change in the distribution of precipitation
a. A brief review on the historical ensemble performances
First, we briefly review the capability of the ensemble to simulate the precipitation and monsoon circulation in Asia. Figure 2a shows the annual signal of precipitation in Asia during the period of 1976 to 2005 (HIST) for the ensemble mean, and GPCP daily data (averaged over 1997–2007). We also computed (for the ensemble and for the observation) the 10th, 90th, and 99th percentiles for each grid point, over the entire period, and then averaged them for the Asia domain (5°–45°N, 65°–160°E).
Annual signal of precipitation (mm day−1) in (a) the Asia region (5°–45°N, 65°–160°E), (b) EAR (22°–45°N, 105°–145°E), (c) IR (5°–28°N, 70°–105°E), and (d) WNPR (5°–22°N, 105°–160°E). The solid line is the GPCP observations, and dashed line is the CMIP5 ensemble mean. For Asia region (left), on the bottom panel, middle lines are the mean precipitation, and upper (lower) lines with cross symbols are the 90th (10th) percentiles. In the top panel, lines indicate the 99th percentile, and triangles (squares) correspond to the mean value of the 10 models with highest (lowest) resolution. The signal is averaged over 1976–2005 for the CMIP5 ensemble, and 1997–2007 for the GPCP observations.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
The mean signal of the ensemble is quite consistent with the observation, but we can notice that the spread of precipitation, indicated by the 10th and 90th percentiles, is underestimated by the CMIP5 ensemble during the summer. This may be partly due to the difference in resolution between observation (1°) and the mean CMIP5 models (most of them having a resolution between 1.5° and 2.5°). This point should be particularly true for highest precipitation. Thus, for the 99th percentile, we also separated the 10 models with the highest resolution (triangles) and the 10 models with the lowest resolution (squares). The impact of the resolution appears clearly, with an improved estimation of the 99th intensity according to the highest-resolution models during the Asian summer monsoon (triangles are closer to the observation line).
Spatially, the precipitation exhibits more differences, as shown in Fig. 3. We identify two periods: May to June (MJ), when the mei-yu front is more active), and July to August (JA), when the monsoon signal strengthens in the IR and WNPR); see Ding (2007) for more details about phases of the Asian summer monsoon. Although the main circulations of the different phases are adequately represented by the ensemble, bias regarding the precipitation remained. The models underestimate heavy and extreme precipitation, especially during the summer (Fig. 2a). The models with a higher resolution (triangles in Fig. 2a) demonstrate a higher ability to simulate correctly the magnitude of extreme events during summer but still exhibit a low bias during winter, compared with GPCP observation. The mean precipitation of the ensemble is more consistent with the observation when averaged over Asia (Fig. 2a), but the subregions also exhibit some bias. In the EAR (Figs. 2b and 3), the models underestimate the precipitation during all of the simulated summer monsoons, with a larger bias during the MJ period. In the IR (Figs. 2c and 3), the onset of the monsoon is likely delayed, leading to a negative bias during MJ, particularly in the Bay of Bengal. The underestimated precipitation resulted from the unrealistic topography in the models, as indicated by Wu et al. (2014). By contrast, during the summer, the models overestimate precipitation over the ocean and underestimate precipitation over land. Finally, in the WNPR, the models overestimate precipitation during the entire year. These biases of the CMIP5 ensemble were studied by Lee and Wang (2014) and could significantly affect the reliability of the projection. Therefore, because of the systematic bias of the ensemble, the results of this study must be considered with a margin of error.
Mean precipitation (shading and contours, mm day−1) and 850-hPa winds (vectors) for (left) GPCP precipitation and NCEP–NCAR winds reanalysis, (middle) CMIP5 ensemble mean, and (right) the difference of precipitation between CMIP5 ensemble mean and GPCP observation, for the (top) May–June (MJ) and (bottom) July–August (JA) periods. All variables are averaged over 1976–2005, except GPCP precipitation, which is averaged over 1997–2007.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
b. Confidence on the projected change
Figure 4 depicts the change in the probability density function (pdf) of precipitation projected by the CMIP5 models, in the three specific regions. It shows, from percentiles ranging from 1 to 99 (every 10 percentiles between 10 and 90 and every percentile between 90 and 99; the first percentile is identified as 0), the relative change between RCP and HIST, for the intensity (left) and frequency (right). On the frequency chart, the change in the frequency of dry days (identified as daily precipitation lower than percentile 1) is shown before the 0 value. The wet and dry seasons are separated. The wet season, defined as the four consecutive months in HIST with the highest precipitation (averaged for each region), is plotted in blue bar charts. The dry season, defined as the four consecutive months in HIST with the lowest precipitation (averaged for each region), is plotted in orange bar charts. The wet and dry season vary for each region: EAR: May to August (wet) and December to March (dry); IR: June to September (wet) and January to April (dry); WNPR: July to October (wet) and January to April (dry). For each bin, we indicate the 25th to 75th ensemble quartile intervals (colored boxes) and the 10th to 90th ensemble quartile intervals (bars). The black horizontal bar inside the color chart shows the 50th ensemble percentile. To extract the most confident signal, we introduced a color scale to mark the values that passed a 90% or 95% Student’s t test.
Probability density function (pdf) of relative change (in percent) in precipitation (left) intensity and (right) frequency. Precipitation is divided into 10 bins (from 1 to 100), and the last 10% is also divided into 10 bins of 1% each. Results are separated between the wet season (blue charts) and dry season (orange charts) of each region. The color boxes show the 25th–75th ensemble quantile, and the bars indicate the 10th–90th ensemble quantile. The black horizontal bars inside the boxes show the 50th ensemble percentile. In the frequency plots, the first bin (before 0) represent the dry days. Color scale indicates the values that passed a certain confidence level (90% or 95%), computed from a Student’s t test.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
The results during the dry season are generally less certain (Fig. 4, orange-red charts), but a seasonal tendency clearly appear in the EAR. The intensity (and, with lower confidence, the frequency) of medium precipitation exhibits a decrease during the dry period. Only the extreme precipitation (98th and 99th percentiles) increases significantly during this period. In the IR, precipitation exhibits a decrease in intensity and frequency, whereas in the WNPR the precipitation tends to increase. During the wet season (Fig. 4, blue charts), the intensity of all precipitation increases in each region. The change in frequency of medium and light precipitation is not clear, although it tends to decrease (confirmed by an increase in dry-days frequency), whereas the frequency of heavy and extreme precipitation keep increasing. Changes in extreme precipitation (99th percentile) are more confident, in every regions, compared with the change of medium and light precipitation. For the EAR and the IR, the 10th ensemble percentile of intensity of extreme precipitation is greater than 0, meaning that most of the models confirm an increase in extreme precipitation during the wet season. This is not the case for the WNPR where ensemble spread is greater. The increases in frequency of extreme events in each region are also higher (35%–40%) than the increase in intensity (10%–20%).
The distinct behaviors between dry and wet season tend to increase the seasonal range of precipitation in the EAR and the IR, possibly correlated with a change in atmospheric dynamics. This question is addressed, for the wet season, in section 4. The uncertainties associated with the ensemble projection, especially in the IR and the WNPR, were indicated by Turner and Slingo (2009). This highlights the difficulty that models demonstrate in simulating precipitation over tropical regions and oceans, and also the range of parameterization scheme between the models, and their response to the radiative forcing. We investigate this point in the next section.
c. Scattering of the ensemble model projection
A part of the uncertainty could be due to the response of the surface temperature to the increase in radiative forcing in the different models. Some models exhibit only a weak increase in temperature, which could explain the lower increase in (extreme) precipitation. To investigate this hypothesis, we plot the changes in extreme precipitation versus the change in surface temperature in each model during the wet season (Fig. 5, top). However, the results do not exhibit a significant correlation between the response in surface temperature in different models and the change in extreme precipitation (intensity or frequency). Thus, other sources of uncertainty may affect the projection, such as the resolution, the representation of the dynamics, and the parameterization of each model. They translate the actual limits of the knowledge of the climate system of the scientific community. The scattering of the ensemble can also be used to confirm several characteristics of the projection. Indeed, an additional aspect that is consistent with previous studies, such as Kharin and Zwiers (2005), Chou and Lan (2012), and Chou et al. (2013), is that the increase of extreme rainfall intensity is greater than the average precipitation. This is illustrated in Fig. 5c, in which the change of median precipitation intensity is compared with the change of extreme precipitation intensity (during the wet season). Most of the points are above the 1–1 relationship line, indicating a faster increase in extreme event intensity. Moreover, the change in the characteristics of precipitation is highlighted when comparing the change in frequency of precipitation (all wet days) with the change in frequency of only extreme precipitation (99th percentile), in Fig. 5d, during the wet season. The negative correlation is obvious, particularly for the EAR (black crosses) and IR (green squares). Thus, precipitation is expected to be slightly less frequent but more intense during the wet season.
Scatterplots of the relative change during the wet season (for each model) in extreme precipitation (y axis) vs (a),(b) surface temperature (°C) and (c),(d) mean precipitation (percent). Black crosses, green squares, and red triangles indicate respectively the EAR, IR, and WNPR. Diagonal line indicates the Clausius–Clapeyron relationship for (a) and (b) and the identity (1–1) relationship for (c) and (d).
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
The previous results indicates various behaviors and reliability (based on our model ensemble) of projected changes in various periods and regions. Clearly, the results are stronger for extreme events, during the wet season in the EAR. In the IR and the WNPR, results are less confident, but still exhibit significant trends for the extreme events, during the wet season. Therefore, we can now study the tendency of extreme precipitation, as well as the associated dynamics, during the wet season (i.e., the summer monsoon) with an adequate confidence.
4. Intensification of extreme during the summer monsoon: A change in atmosphere dynamics
The Asian summer monsoon is characterized by active phases and short breaks. However, given the time scale of the projection, and because we focus on the general tendency of extreme precipitation, we summarize the results during two averaged periods: May to June (late spring) and July to August (summer). Figure 6 shows the intensity of percentile 99 (pct99) for HIST and RCP, as well as the difference. Figure 7 shows the same values for fqpct99. For both figures indicating differences, only the regions that passed a 90% confidence Student’s t test are shaded.
Intensity of extreme precipitation (percentile 99), for (left) HIST, (middle) RCP, and (right) the difference RCP − HIST. For HIST and RCP, contours are plotted every 10 mm day−1. On the difference plots, contours are plotted every 5 mm day−1 and the shade values passed a 90% confidence level with Student’s t test. All values are in mm day−1.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
As in Fig. 6, but for the frequency of the 99th percentile. All values are in percentage of days. Contours are plotted every 2% for HIST and RCP, and every 0.5% for the difference plots.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
The mean circulation in low-level atmosphere (850 hPa) is presented in the next section based on its change in monthly moisture flux (MF; Figs. 8a,b) and its horizontal convergence (HMFC; Fig. 9). Figures 8c and 8d depict the change in the circulation at 850 hPa. The monthly-mean atmospheric condition may not be suitable for studying extreme events, because such events are often associated with specific and local conditions. In particular, during the Asian summer monsoon, tropical cyclones (TCs) play a major role in triggering extreme events. However, the climate models exhibit an unfavorable spatial resolution and may not correctly simulate the TCs. Therefore, in section 4a, we describe the mean change of the atmosphere as a background signal of the environment that could favor the formation of extreme precipitation, independently of the ability of models to simulate phenomena such as TCs. In sections 4b and 4c we describe in detail the horizontal and vertical moisture convergence changes during the extreme events, implicitly including the TCs.
Change in 850-hPa moisture flux (left, g kg−1 m s−1) and winds (right, m s−1), averaged over (top) MJ and (bottom) JA. For the winds, contours are plotted every 0.5 m s−1, and vectors are plotted only for values greater than 0.5 m s−1. Scales of vectors are indicated on the top-left corner of each panel (20 g kg−1 m s−1 for the moisture flux, 1 m s−1 for the winds). For the moisture flux, shade values passed a 90% confidence level with Student’s t test.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
Change in 850-hPa horizontal moisture flux convergence (HMFC; g kg−1 s−1 10−5). The (a),(d) total HMFC is separated into its (b),(e) convergence and (c),(f) advection parts. Only the positive component of the moisture flux convergence (based on values from the total background field HIST) is plotted. Thus, positive (negative) anomalies in (a) and (d) indicate an increase (decrease) of the total background signal. The values are averaged over (top) MJ and (bottom) JA. Positives values are contoured with full lines, and negative values with dashed lines.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
a. Mean low-level circulation and moisture flux convergence
In this paper, we express HMFC as the sum of the contributions of a convergence 
1) May to June
We first consider the period from May to June [Figs. 6, 7, and 9 (top) and Figs. 8a,c].
In Figs. 6 and 7, the mei-yu front (EAR), which develops during late spring, is clearly exhibiting an intensification in both frequency and intensity of extreme rainfall. This tendency is correlated with an increase in the MF (Fig. 8a) bringing more low-level humidity from the south along the East Asian coast. By contrast, the wind (Fig. 8b) does not exhibit a clear tendency in this region (except a slight increase). The change in MF is mostly attributable to a change in the atmospheric humidity. The total HMFC also increases (Fig. 9a). Convergence and advection seem to play an equal role during this period (Figs. 9b,c). They both strengthen the total HMFC.
In other regions, the tendency for changes in extreme events is not confident (Figs. 6c and 7c). In the WNPR, a tendency of increase in extreme frequency starts to develop, indicating a clear increase in HMFC in this region (Fig. 9a) and dominated by the convergence part (Fig. 9b). In the IR, despite an increase in HMFC (Fig. 9a), the change in extreme events (frequency or intensity) is unclear during this period.
2) July to August
In the following we present summer period, July to August [Figs. 6, 7, and 9 (bottom) and Figs. 8b,d].
In the EAR, the tendency in frequency and intensity of extreme events is similar with the MJ period, although the confidence for the intensity change decreases in the south part (Figs. 6f and 7f). The increase in MF occurs in a large region, especially over the Korean Peninsula and eastern China (Fig. 8b), which was also indicated by Min et al. (2012). The change of the associated circulation (Fig. 8d) is unclear, particularly over the ocean. We can still notice an increase in northward winds in eastern China, bringing more moisture over the continent. HMFC also exhibits an increase in the north part of the continent (Fig. 9d), but the magnitude of change is slightly reduced over the ocean. This is attributable to the fact that during this period the advection part tends to decrease, whereas the convergence part increases (Figs. 9e,f), so both contributions oppose each other with a similar magnitude. This is different from the MJ period when both contributions exhibit the same sign.
In the IR, the change in extreme events becomes significant (Figs. 6f and 7f), especially the intensity over the continent. The ocean region remains uncertain. The modification of the associated circulation in the region exhibits a strong signal (Figs. 8b,d): the westerlies increase over the north part, associated with an increase of MF, and decrease in the south part, associated with a (not confident) decrease of MF. This difference of behavior between the northern and southern IR suggest that the monsoon system in this region could move northward. Nevertheless, HMFC increases over the entire region (Fig. 9d), largely dominated by its convergence part (Fig. 9e). Thus, the convergence of moisture over the entire region provides more humidity to favor extreme events during the mature phase of the monsoon in the IR.
In the WNPR, the change in intensity (Fig. 6f) remains uncertain, but the increase in frequency (Fig. 7f) is confident over a large region. The associated MF does not exhibit a confident increase in this region (Fig. 8b), but the westerlies exhibit an increase in the west part, and a decrease in the east part (Fig. 8d). It can be considered as a deepening of the monsoon trough in the WNPR. During this period, the HMFC exhibits a clear increase, dominated by the convergence term (Figs. 9d,e).
The change in the mean low-level circulation and MF provides a rough idea of the processes that can favor extreme events. In the EAR, mostly during MJ, both HMFCa and HMFCc lead to an increase in moisture convergence, while both intensity and frequency of extremes increase. In the IR, HMFC mostly increases during JA (controlled by HMFCc), with a substantial increase in the intensity of extremes. The change of winds suggests that the India monsoon moves northward. In the WNPR, the increase in HMFC is the most significant during summer, with a deepening of the monsoon trough and an increase in the frequency of extremes. The increase in low-level humidity over tropical ocean regions provides the necessary elements to favor the increase in extremes. These changes in circulation tend to increase the specific humidity in the low level of the atmosphere, providing more potential fuel to trigger deep and intense convection. They vary in time and space, as do the changes in precipitation, and therefore constitute a probable factor for the spatial and time variation in the behavior of extreme events.
Nevertheless, extreme events are rare by definition and associated with local and specific conditions that can differ significantly from the mean state. To further investigate the contributions of circulation and moisture during extreme precipitation, we study the associated daily HMFC (with HMFCa and HMFCc) and present the findings in the following section.
b. Daily horizontal moisture flux convergence during extreme events
In the following two sections, we only use 22 models (because of the availability and storage of full 3D daily data), marked with an asterisk in Table 1.
For this part of the study, we retain the same definition of HMFC, HMFCa and HMFCc, as defined in the previous section; however we compute the terms only at the time and location of each extreme event (99th percentile). This means that for each grid point and day that is considered an extreme event (i.e., precipitation that exceeds the threshold value), we compute the moisture flux convergence (at this special point and day). Other points are masked. We then average (for each model) the terms over a region (EAR, IR, and WNPR) and compute the monthly mean over the 30 years of HIST and RCP. For each model we obtain the monthly signal for each region, based on daily computation only during extreme events. Subsequently, we compute the ensemble mean change (between RCP and HIST) and the associated standard deviation. The results are summarized in Table 2, in which the ensemble mean change (in percent) is indicated for each region and for the two periods (MJ and JA), and its associated standard deviation is specified in brackets. Results with a mean greater than the standard deviation are in bold type face.
Relative change (in percent) in HMFC (at 850 hPa) during extreme events (99th percentile), during the wet season, separated into its convergence and advection parts. We indicate the ensemble mean and the ensemble standard deviation between brackets. Bold text indicates cases were the mean is greater than the standard deviation.
The first point to notice is the high uncertainties associated with the changes. In almost every case, the standard deviation is larger than the mean change. This is particularly true for the advection term. The confidence for the convergence term and the total HMFC is not high either but can still be considered. The uncertainties are generally lower for the EAR, especially for the convergence term. Thus, the models agree more favorably regarding the change of the atmospheric state in this region during extreme events. The total changes of HMFC in this region (11% and 4%) correspond obviously to the convergence term (14% and 9%) reduced by the advection term (1% and −11%). Although less certain, HMFCa plays a significant role in this region, limiting HMFC during extreme events. By contrast, the total changes in IR (10% and 11%) and WNPR (17% and 18%) are very close to the convergence terms (10% and 11% for IR, and 18% and 18% for the WNPR), demonstrating the dominance of this contribution. In addition, the increase in HMFC is the highest in the WNPR and the lowest in the EAR. This indicates the variation in changes between tropical and subtropical regions, with the atmospheric moisture increasing considerably faster in tropical regions. According to these results, the change in HMFC during extreme events increases during MJ, and is mostly determined by an increase in HMFCc.
Even if the ensemble mean is low compared with the standard deviation, it is possible to use the scattering of the ensemble to study the correlation between the change in extreme events and the change in HMFC. For each model, the mean change in pct99 and fqpct99, as well as the corresponding change in HMFC, HMFCc, and HMFCa (computed only during extreme events), is averaged in each region and during each period. Then, for each model, we have (for each period, region, and variable) one couple (e.g., the mean change in pctl99 vs the mean change in HMFC, in the EAR, during MJ). The ensemble results provide 22 couples, which we use to compute a linear correlation. The purpose is to determine whether a change in extremes can be easily explained by a change in HMFC. Because this correlation is computed using an ensemble, it can also be interpreted as an explanation of the scattering of the ensemble (i.e., whether a variation in change of MFC explains the difference in change of precipitation). Each model involves a specific parameterization, which clearly leads to a lower correlation between them. Moreover, the linear correlation is a first approximation and does not account for all nonlinear effects. The results are summarized in Table 3, with values in bold type face indicating correlation coefficients greater than 0.40.
Linear correlation coefficients between the relative change in extreme precipitation and the relative change in HMFC (at 850 hPa) during the extreme events (99th percentile), during the wet season, separated into its convergence and advection parts. Bold text indicates values higher than 0.40.
The results exhibit higher correlation between pct99 and HMFC (i.e., between the intensity of extremes and the total HMFC), and during JA, with coefficients larger than 0.6 (0.68 for the EAR, 0.64 for the IR, and 0.64 for the WNPR). These coefficients are close to the HMFCc correlation (0.69 for the EAR, 0.66 for the IR, and 0.62 for the WNPR), indicating the importance of this term on the total HMFC modification. The correlation coefficients are generally higher for pctl99 than for fqpct99, except in the IR during MJ. The horizontal MFC exerts a greater influence on the intensity of the extremes than on their frequency. For the EAR (in contrast to the previous section, which that suggests a possible role of HMFCa in the change of extreme events), the correlation suggests that this term has an insubstantial influence. The coefficients are considerably low (−0.01 and 0.01 for pct99, −0.28 and 0.13 for fqpct99), as they are for other regions. Thus, even if HMFCa can influence HMFC when we consider monthly mean (Fig. 9), it does not seem to explain the characteristics of change in extreme events. In the IR, fqpct99 is significantly influenced by HMFC, particularly during MJ. This means that during this period, which corresponds to the onset of the monsoon in this region, the change in extreme events is significantly influenced by the low-level circulation change. Moreover, because models may not correctly simulate this onset, the confidence of projection of fqpct99 (Fig. 7c) might be low, compared with other regions during the same period. However, pctl99 exhibits lower correlations during the same period. Thus, it is less influenced by the uncertainties during the onset. In the WNPR, correlations are over 0.5 for each period (MJ and JA) for pctl99. Hence, in this region, the horizontal MFC (governed by its convergence term) in the lower atmosphere is a major factor determining the intensity of extreme events during the summer, and not only JA. This signal is consistent with the monthly mean signal (Fig. 9).
The horizontal convergence also occurs at higher levels in the troposphere, and may play a role in the precipitation change. To check this point, we computed the correlation as in Table 3, but this time we use the 100–1000-hPa integration of HMFC (and its two contributions). Results are indicated in Table 4. Correlations with the column-integrated horizontal MFC are similar to or lower than the correlation with the 850-hPa level. This highlights the dominant role of the low level circulation and convergence, at least to explain the scattering of the ensemble. The column integration of the horizontal MFC does not bring more information to explain the variation between models.
As in Table 3, but for the column-integrated (1000–100 hPa) HMFC during the extreme events (99th percentile).
Regarding the correlation coefficients in the EAR, during MJ, the fact that they are low (0.23 for pctl99 and −0.3 for fqpct99) can be attributed to the low scattering of the ensemble. As shown in Table 2, the ensemble standard deviation in the EAR during MJ (compared with the ensemble mean change) is the lowest (10% for HMFC and HMFCc). When the dispersion of the ensemble is low, the linear correlation becomes highly sensitive (finding a linear direction is more difficult in a small potato-shaped ensemble than in a well-dispersed ensemble, assuming the scattering is organized in a linear shape). Thus, even if the correlation between HMFC and the extreme precipitation is low in the EAR during MJ, it does not mean that the correlation does not exist. The ensemble method is just not able to detect the correlation, because we used the ensemble dispersion to compute the correlations (i.e., if the ensemble dispersion is too weak, a clear linear correlation may not appear). Therefore, we cannot conclude whether this correlation is significant.
c. Daily vertical moisture flux convergence during extreme events
If the low-level circulation and moisture convergence can provide the required moisture for precipitation, the vertical structure of the atmospheric specific humidity and circulation in the troposphere is also important for extreme events. In this study, we use a simple definition of the vertical moisture advection (VMFC), defined by 〈ω∂pq〉, with ω being the vertical velocity, ∂pq repesenting the vertical gradient of specific humidity, and 〈〉 indicating the vertical integration (850–100 hPa); 〈ω∂pq〉 is the most dominant term of the column-integrated moisture budget, especially for heavy precipitation. Similar to the daily HMFC, VMFC is computed only during extreme events (i.e., in a specific place and time of each extreme event), and then averaged per month, in each region, and for each model. Finally, the ensemble mean change and standard deviation are computed. We also separate the two terms ω and ∂pq to analyze their respective contributions. During each extreme event, we compute the associate ω and ∂pq separately (vertically integrated), and compare the mean value of these terms between RCP and HIST. The variable ω represents the dynamical contribution and is denoted VMFCd, and ∂pq represents the thermodynamic part, denoted VMFCt. Finally, we compute the vertical profile (between 1000 and 100 hPa) of VMFC during extreme events (i.e., ω∂pq at each vertical level, and the separated terms VMFCd and VMFCt). The mean integrated change and standard deviation are summarized in Table 5 (bold values indicate cases in which the mean is greater than the standard deviation), and the profiles of changes are shown in Fig. 10. Because the profiles are quite similar between MJ and JA, we plot only the average May to August profiles. In the latter figure, in addition to the ensemble mean (full red line), we plot each separated model (black lines). We also indicate the change in the mean atmosphere (dashed red line).
Relative change (in percent) in VMFC (integrated between 850 and 100 hPa) during extreme events (99th percentile), during the wet season, separated into its dynamical and thermodynamical parts. We indicate the ensemble mean and the ensemble standard deviation between brackets. Bold text indicates cases were the mean is greater than the standard deviation.
Relative change (in percent) of (top) the VMFC profile and its (middle) dynamical and (bottom) thermodynamical parts computed during extreme events (99th percentile). Each change is normalized by its mean historical value of the total atmospheric column (850–100 hPa). Black lines indicate separate models, the full red line is the ensemble mean, and the dashed red line is the ensemble mean of the total atmosphere change (not only during extreme events). Vertical axis indicates the pressure levels (in hPa). The scale is similar for each caption.
Citation: Journal of Climate 28, 4; 10.1175/JCLI-D-14-00449.1
In contrast to HMFC, the changes in VMFC during extremes are considerably clear (Table 5). The total VMFC increases in each region and during each period (36%–47%). VMFCt exhibits the same behavior, with high confidence (because of the very expected increase in moisture in a warmer climate) and a higher magnitude (44%–50%). In contrast, VMFCd is less certain, with a standard deviation greater than the mean change, except in the EAR during MJ. Moreover, the change in VMFCd is negative (−3% to −8%), so it tends to oppose the positive change in VMFCt. The changes are pretty stable between MJ and JA, although VMFC exhibits a slight increase in the IR (36%–43%). In this region, the change in VMFCt is stable between the two periods, and the modification of VMFC is explained by a less negative value of VMFCd (−8% to −3%). In summary, for each region and period, the vertical advection increases during extreme events, mostly because of the change in atmospheric moisture content.
Even if the mean changes are close between the various regions, the vertical profiles of changes exhibit several notable differences (Fig. 10). The VMFC profiles of the models in the EAR (Fig. 10a) are highly consistent; most exhibit a high increase at 500 hPa (30% in mean). By contrast, in the IR and WNPR (Figs. 10b,c), the scattering of the models is greater. The increase in VMFC may occur at lower levels (with a maximum at 500 hPa), but the scattering makes this consideration uncertain. The difference between the change during extreme (full red line) and the mean atmospheric change (dashed red line) is also larger in the IR and WNPR. In the IR, the mean change is larger in the lower atmosphere, whereas in the WNPR it is larger in the entire column. However, these values are relative to their historical mean (as a percentage). Thus, the absolute change of VMFC during extremes is still larger than the absolute mean change (not shown in the figure). VMFCd profiles (Figs. 10d–f) exhibit the same shape in the lower to midtroposphere, with the largest decrease occurring between 850 and 400 hPa (−20% to −25%). In the upper troposphere (above 250 hPa), the change in velocity becomes slightly positive (5%) for all regions (Figs. 10d–f). extreme events. In the WNPR (Fig. 10f), the change in the upper troposphere is more uncertain. VMFCt exhibit the same shape in all regions, with a higher increase in the midtroposphere (35%–50%). The increase is more significant (compared with the mean atmosphere) above 700 hPa. This highlights the importance of the increase of humidity in the upper part of the atmosphere during extremes. Below 700 hPa, the change of VMFCt during extreme events is lower than the change of the mean atmosphere. However, again, these changes are expressed as a percentage of the historical value, and the absolute value of VMFCt during extremes is higher than the absolute mean change (not shown in the figure). In summary, during extreme events, the most significant change is the increase of midtroposphere VMFC (correlated with VMFCt, and moderated by VMFCd).
Similar to HMFC, we computed the linear correlation between the changes in extreme events and the changes in VMFC, VMFCd, and VMFCt. The results are summarized in Table 6.
Linear correlation coefficients between the relative change in extreme precipitation and the relative change in VMFC (integrated between 850 and 100 hPa) during the extreme events (99th percentile), during the wet season, separated into its dynamical and thermodynamical parts. Bold text indicates values higher than 0.40.
In the EAR, the correlation with VMFC is superior during JA, for pctl99 and fqpct99 (0.52 and 0.48, respectively). Coefficients during MJ are nonsignificant (0.29 and −0.1). This may be partially attributable to a too favorable agreement of the models in the changes during this period (as explained in section 4b for HMFC). During in JA, the total correlation increases for both pctl99 and fqpct99 (0.52 and 0.48, respectively). During this period, separated terms exhibit a low correlation with pctl99 (0.34 for VMFCd and 0.32 for VMFCt), while fqpct99 exhibits a greater correlation with VMFCt (0.48) compared with VMFCd (0.10). This implies that the moisture content of the atmosphere in this region significantly influences the triggering of extreme events (fqpct99), whereas the strength of these events (pctl99) depends on both contribution of VMFCt and VMFCd. Furthermore, a switching of the dominant effect between the two periods: during MJ, VMFCt dominates pct99 and VMFCd dominates fqpct99; during JA, VMFCt dominates fqpct99 and VMFCd equally contribute to pct99.
In the IR, the correlation of VMFC is significant for both fqpct99 and pct99, and each period. By contrast, the correlation of VMFCt with pctl99 is considerably low (−0.09 in MJ, 0.06 in JA) compared with VMFCd (0.52 in MJ, 0.80 in JA). Thus, in this region, the strength of extreme events is mostly influenced by the change in vertical velocity. In other words, the scattering of the ensemble change of pctl99 in this region is mostly attributable to the change of dynamics in the models. However fqpc99 exhibits correlations of similar order for both VMFCt and VMFCd, meanings that both terms can significantly influence the triggering of extremes events.
In the WNPR, pct99 exhibits the same characteristics as in the IR (i.e., it is strongly influenced by VMFCd). By contrast, fqpct99 exhibits stronger correlation coefficients with VMFCt. Thus, the change in atmospheric moisture content is crucial in triggering extreme events in this region.
The correlation between VMFC and the intensity of extreme events (pctl99) clearly shows a difference of behavior in the EAR compared with that in the IR and WNPR. In the first region, thermodynamic contribution is dominant during MJ, and both contributions seem crucial during JA, whereas in the two other regions the dynamical term is the main contribution during the two periods. However, the frequency of extreme precipitation is correlated with both contributions in every region in the EAR and IR (with an increase of the role of VMFCt in the EAR during JA), whereas it is dominated by the thermodynamic part in the WNPR. These results must to be considered regarding the results shown in Table 5. The models agree on the change of moisture content in the atmosphere, and thus VMFCt. It is likely to increase the frequency of extremes in every region, with a major contribution in the WNPR, and in the EAR during JA. However, significant correlation with the change in intensity cannot be confirmed, except in the EAR during MJ. The change in vertical velocity during extremes is less confident with this ensemble, except in the EAR; however it is significantly correlated to the change in pctl99 in the IR and WNPR, and (with a lower correlation) with fqpct99 in the EAR. Thus, the change in circulation has a major role for the change in pctl99 (and fqpct99 in the EAR and IR during MJ).
These results may first surprising. An increase in atmospheric humidity content can be expected to be the dominant factor in explaining the change in extreme precipitation. We demonstrated that the dynamical part of the vertical MFC can significantly influence the intensity (and frequency in the EAR and IR during MJ) of these events. Furthermore, because its change is less clear (Table 5), it influences the confidence on the projection of extremes. By contrast, the change in VMFCt is more confident, and should mostly influence the projection of frequency of extremes (and intensity in the EAR during MJ).
d. Summary
In the following section, results from sections 4b and 4c have been combined to provide several conclusions.
The change in low-level circulation (HMFC) is expected to influence (with a positive correlation; Table 3) mostly the intensity of extremes during JA, except in the WNPR where the correlations are significant during MJ. In the IR, it also exerted a significant influence on fqpct99 during both periods. However, the mean change of HMFC and its component (Table 2) in each region exhibits a low confidence. Thus, it could influence (negatively) the confidence of the projected change in extreme events.
On other hand, the vertical moisture convergence, VMFC (Table 6), exhibits a generally favorable correlation with both the intensity and frequency of extreme precipitation (except in the EAR during MJ). Its thermodynamic term (VMFCt) exhibit a higher correlation with frequency, and the dynamical part influences mostly the intensity (and fqpct99 in the IR and EAR, with correlations of approximately the same level of VMFCt). Thus, the change in moisture would mostly increase the frequency of extremes, but the change in local dynamics (vertical winds) would be the main driver of the intensity of these events in the WNPR, and both contributions could influence fqpct99 in the EAR and IR. The mean change in VMFC is more certain (Table 5), particularly VMFCt, but the two terms have opposed signs (VMFCd is negative, whereas VMFCd is positive). The VMFCd trend tends to be negative and thus moderates the increase in pct99 and fqpct99.
In a warming climate, it is generally accepted that the specific humidity of the atmosphere would increase. Hsu et al. (2012) showed that the variation in monsoon areas and precipitation is linked to the increase of moisture convergence, offset by the change in circulation. We demonstrated that the circulation component of the vertical MFC exerts potentially high influence on trends of extreme events, especially the intensity of these events in the IR and WNPR. However, its influence varies significantly between regions (in the EAR, the thermodynamic component is also crucial), which is consistent the findings of Chen et al. (2012) demonstrating that the dynamic contribution induces spatial variation of changes. We also demonstrated that the influence of each component varies significantly according to time, comparing the onset period (MJ) and summer period (JA). This is particularly so for the influence of HMFCc on intensity (i.e., the circulation) that increases significantly between MJ and JA (Table 3). Furthermore, the confidence of the ensemble projected change of extreme precipitation may be influenced by the low agreement of model dynamics, especially in the IR and WNPR.
The above considerations have been obtained with a simple definition of moisture flux convergence. Residual terms that are not explicitly included and analyzed could play a role in the formation of extremes. Moreover, the ability of various climate models to simulate phenomena such as TCs, may introduce bias in ensemble statistics. Thus, correlations presented in this paper should be considered of limited validity, as indicators explaining what can influence the projection of extreme precipitation, and consequently, what can influence the ensemble scattering and uncertainties for these projections.
5. Concluding remarks
In this study, we analyzed the tendency in precipitation change during the Asian summer monsoon, especially extreme changes, using 30 models from a CMIP5 ensemble with the RCP 8.5 pathway. Because of the ensemble averaging, many uncertainties emerged; consequently, only some of the results can clearly be identified with high confidence (Fig. 4). We particularly emphasized the confidence of the projections, and the link with the modification of the atmospheric dynamics. The ensemble method enabled identifying the most consistent projections, and the dispersion of the ensemble was used to compute correlations between the change of extreme precipitation and the circulation change.
Based on the selected models, there is high confidence that the East Asia region would be highly vulnerable to global warming, exhibiting more intense and more frequent extreme precipitation. The India region also exhibits the same tendency, but with less confidence. The results in the western North Pacific region are less certain results, particularly regarding the intensity of extremes. These variations in behavior highlight the importance of considering regional investigation for monsoon changes, as Wang and Ding (2006) showed by considering land and ocean monsoon areas in their study.
The changes in monthly mean circulation and moisture flux convergence (Figs. 8 and 9) during the summer monsoon indicate the spatial and temporal characteristics in each region. We were mainly able to clarify the findings by considering the moisture convergence computed only during extreme events (Fig. 10 and Tables 2–6). We demonstrated that the vertically integrated moisture gradient can influence the frequency of extreme events, but the intensity of these events is strongly correlated with the change of HMFC in low-level and dynamical term of VMFC. The agreement of models regarding the change in circulation and VMFCd being low may explain the uncertainties associated with the projection (Fig. 4).
Thus, the increase in atmospheric moisture content serves to strengthen the precipitation, but the change in characteristics of extremes should be considered with particular attention to the (spatially and temporally) local variations of the circulation.
These results have to be considered in a large context of uncertainties. The systematic bias of the CMIP5 models (indicated in section 3), the low ability of models to simulate tropical cyclones correctly, and the definition of the moisture flux we adopted could lead to a biased estimation of the correlations we computed. Thus, the correlations presented in this paper should be considered of validity extend and as indicators explaining what can influence the projection of extreme precipitation and, consequently, what can influence the ensemble dispersion and uncertainties for these projections. Finally, in this study, we considered each model with equal weight to compute the ensemble statistics. To increase the confidence of the results and the consistency of the ensemble mean, an alternative method would be to weight or select each model based on its historical performance to simulate the summer monsoon signal in Asia. However, for this study, we adopted a large panel of models, regardless of their performance, and used ensemble scattering to compute correlations and the confidence of projections.
Acknowledgments
This study was supported by the National Science Council, Taiwan, under Grant NSC-100-2119-M-001-029-MY5. We thank the three anonymous reviewers who gave many leads to complete our study and helped to improve the quality of this paper.
REFERENCES
Barnett, D. N., S. J. Brown, J. M. Murphy, D. M. H. Sexton, and M. J. Webb, 2006: Quantifying uncertainty in changes in extreme event frequency in response to doubled CO2 using a large ensemble of GCM simulations. Climate Dyn., 26, 489–511, doi:10.1007/s00382-005-0097-1.
Chang, C.-P., B. Wang, and N.-C. G. Lau, 2005: The global monsoon system: Research and forecast. WMO/TD 1266, TMPR Rep. 70, 542 pp.
Chen, C.-A., C. Chou, and C.-T. Chen, 2012: Regional perspective on mechanisms for tropical precipitation frequency and intensity under global warming. J. Climate, 25, 8487–8501, doi:10.1175/JCLI-D-12-00096.1.
Chou, C., and J. D. Neelin, 2004: Mechanism of global warming impacts on regional tropical precipitation. J. Climate, 17, 2688–2701, doi:10.1175/1520-0442(2004)017<2688:MOGWIO>2.0.CO;2.
Chou, C., and C.-W. Lan, 2012: Changes in the annual range of precipitation under global warming. J. Climate, 25, 222–235, doi:10.1175/JCLI-D-11-00097.1.
Chou, C., J. D. Neelin, C.-A. Chen, and J.-Y. Tu, 2009: Evaluating the “rich-get-richer” mechanism in tropical precipitation change under global warming. J. Climate, 22, 1982–2005, doi:10.1175/2008JCLI2471.1.
Chou, C., C.-A. Chen, P.-H. Tan, and K.-T. Chen, 2012: Mechanisms for global warming impacts on precipitation frequency and intensity. J. Climate, 25, 3291–3306, doi:10.1175/JCLI-D-11-00239.1.
Chou, C., J. C. H. Chiang, C.-W. Lan, C.-H. Chung, Y.-C. Liao, and C.-J. Lee, 2013: Increase in the range between wet and dry season precipitation. Nat. Geosci., 6, 263–267, doi:10.1038/ngeo1744.
Christensen, J. H., and Coauthors, 2007: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 847–940.
Ding, Y., 1994: Monsoon over China. Kluwer Academic, 419 pp.
Ding, Y., 2007: The variability of the Asian summer monsoon. J. Meteor. Soc. Japan, 85B, 21–54, doi:10.2151/jmsj.85B.21.
Duan, A., J. Hu, and Z. Xiao, 2013: The Tibetan Plateau summer monsoon in the CMIP5 simulations. J. Climate, 26, 7747–7766, doi:10.1175/JCLI-D-12-00685.1.
Giorgi, F., E.-S. Im, E. Coppola, N. S. Diffenbaugh, X. J. Gao, L. Mariotti, and Y. Shi, 2011: Higher hydroclimatic intensity with global warming. J. Climate, 24, 5309–5324, doi:10.1175/2011JCLI3979.1.
Hsu, P.-C., T. Li, J.-J. Luo, H. Murakami, A. Kitoh, and M. Zhao, 2012: Increase of global monsoon area and precipitation under global warming: A robust signal? Geophys. Res. Lett.,39, L06701, doi:10.1029/2012GL051037.
Hsu, P.-C., T. Li, H. Murakami, and A. Kitoh, 2013: Future change of the global monsoon revealed from 19 CMIP5 models. J. Geophys. Res. Atmos., 118, 1247–1260, doi:10.1002/jgrd.50145.
Huffman, G., R. F. Adler, M. M. Morrissey, D. T. Bolvin, S. Curtis, R. Joyce, B. McGavock, and J. Susskind, 2001: Global precipitation at one-degree daily resolution from multisatellite observations. J. Hydrometeor., 2, 36–50, doi:10.1175/1525-7541(2001)002<0036:GPAODD>2.0.CO;2.
Inoue, T., and H. Ueda, 2011: Delay of the first transition of Asian summer monsoon under global warming condition. SOLA,7, 081–084, doi:10.2151/sola.2011-021.
Jones, C., and L. M. V. Carvalho, 2013: Climate change in the South American monsoon system: Present climate and CMIP5 projections. J. Climate, 26, 6660–6678, doi:10.1175/JCLI-D-12-00412.1.
Kharin, V. V., and F. W. Zwiers, 2005: Estimating extremes in transient climate change simulations. J. Climate, 18, 1156–1173, doi:10.1175/JCLI3320.1.
Kripalani, R. H., J.-H. Oh, and H. S. Chaudhari, 2007: Response of the East Asian summer monsoon to doubled atmospheric CO2: Coupled climate model simulations and projections under IPCC AR4. Theor. Appl. Climatol., 87, 1–28, doi:10.1007/s00704-006-0238-4.
Kusunoki, S., and O. Arakawa, 2012: Change in the precipitation intensity of the East Asian summer monsoon projected by CMIP3 models. Climate Dyn., 38, 2055–2072, doi:10.1007/s00382-011-1234-7.
Lee, J.-Y., and B. Wang, 2014: Future change of global monsoon in the CMIP5. Climate Dyn., 42, 101–119, doi:10.1007/s00382-012-1564-0.
Ma, J., S.-P. Xie, and Y. Kosaka, 2012: Mechanisms for tropical tropospheric circulation change in response to global warming. J. Climate, 25, 2979–2994, doi:10.1175/JCLI-D-11-00048.1.
Meehl, G. A., J. M. Arblaster, and C. Tebaldi, 2005: Understanding future patterns of increased precipitation intensity in climate model simulations. Geophys. Res. Lett.,32, L18719, doi:10.1029/2005GL023680.
Meehl, G. A., and Coauthors, 2007: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–845.
Min, S.-K., S. Legutke, A. Hense, U. Cubasch, W.-T. Kwon, J.-H. Oh, and U. Schlese, 2006: East Asian climate change in the 21st century as simulated by the coupled climate model ECHO-G under IPCC SRES scenarios. J. Meteor. Soc. Japan, 84, 1–26, doi:10.2151/jmsj.84.1.
Min, S.-K., S. Legutke, A. Hense, U. Cubasch, W.-T. Kwon, J.-H. Oh, and U. Schlese, 2012: Projected changes in Asian summer monsoon in RCP scenarios of CMIP5. Atmos. Oceanic Sci. Lett., 5, 43–48.
Räisänen, J., 2005: Impact of increasing CO2 on monthly-to-annual precipitation extremes: Analysis of the CMIP2 experiments. Climate Dyn., 24, 309–323, doi:10.1007/s00382-004-0510-1.
Ramage, C. S., 1971: Monsoon Meteorology. International Geophysical Series, Vol. 15, Academic Press, 296 pp.
Riahi, K., A. Gruebler, and N. Nakicenovic, 2007: Scenarios of long-term socio-economic and environmental development under climate stabilization. Technol. Forecasting Soc. Change, 74, 887–935, doi:10.1016/j.techfore.2006.05.026.
Scoccimarro, E., S. Gualdi, A. Bellucci, M. Zampiery, and A. Navara, 2013: Heavy precipitation events in a warmer climate: Results from CMIP5 models. J. Climate, 26, 7902–7911, doi:10.1175/JCLI-D-12-00850.1.
Seager, R., N. Naik, and G. A. Vecchi, 2010: Thermodynamic and dynamic mechanisms for large-scale changes in the hydrological cycle in response to global warming. J. Climate, 23, 4651–4668, doi:10.1175/2010JCLI3655.1.
Seo, K.-H., J. Ok, J.-H. Son, and D.-H. Cha, 2013: Assessing future change in the East Asian summer monsoon using CMIP5 coupled models. J. Climate, 26, 7662–7675, doi:10.1175/JCLI-D-12-00694.1.
Seth, A., S. A. Rauscher, M. Biasutti, A. Giannini, S. J. Camargo, and M. Rojas, 2013: CMIP5 projected changes in the annual cycle of precipitation in monsoon regions. J. Climate, 26, 7328–7351, doi:10.1175/JCLI-D-12-00726.1.
Shiu, C.-J., S. C. Liu, C. Fu, A. Dai, and Y. Sun, 2012: How much do precipitation extremes change in a warming climate? Geophys. Res. Lett., 39, L17707, doi:10.1029/2012GL052762.
Stephens, G. L., and T. D. Ellis, 2008: Controls of global-mean precipitation increases in global warming GCM experiments. J. Climate, 21, 6141–6155, doi:10.1175/2008JCLI2144.1.
Tebaldi, C., K. Hayhoe, M. Arblaster, and G. A. Meehl, 2006: Going to the extremes. Climatic Change, 79, 185–211, doi:10.1007/s10584-006-9051-4.
Trenberth, K. E., A. Dai, R. Rasmussen, and D. Parsons, 2003: The changing character of precipitation. Bull. Amer. Meteor. Soc., 84, 1205–1217, doi:10.1175/BAMS-84-9-1205.
Turner, A. G., and J. M. Slingo, 2009: Uncertainties in future projections of extreme precipitation in the Indian monsoon region. Atmos. Sci. Lett., 10, 152–158, doi:10.1002/asl.223.
Turner, A. G., and H. Annamalai, 2012: Climate change and the South Asian summer monsoon. Nat. Climate Change, 2, 587–595, doi:10.1038/nclimate1495.
Wang, B., 2006: The Asian Monsoon. Praxis Publishing, 787 pp.
Wang, B., and Q. Ding, 2006: Changes in global monsoon precipitation over the past 56 years. Geophys. Res. Lett.,33, L06711, doi:10.1029/2005GL025347.
Wu, C.-H., and M.-D. Chou, 2012: Upper-tropospheric forcing on late-July monsoon transition in East Asia and western North Pacific. J. Climate, 25, 3929–3941, doi:10.1175/JCLI-D-11-00343.1.
Wu, C.-H., W.-S. Kau, and M.-D. Chou, 2009: Summer monsoon onset in the subtropical western North Pacific. Geophys. Res. Lett.,36, L18810, doi:10.1029/2009GL040168.
Wu, C.-H., H. H. Hsu, and M.-D. Chou, 2014: Effect of the Arakan Mountains in the northwestern Indochina Peninsula on the late May Asian monsoon transition. J. Geophys. Res. Atmos., 119, 10 769–10 779, doi:10.1002/2014JD022024.
