Diurnal cycle is one of the most fundamentalmodes of climatic variability driven by the solar radiation. Diurnal signals are discernible in many meteorological variables, such as the temperature, wind, pressure, and precipitation. Affected by the inhomogeneous surface forcing formed by the l and -sea contrast and the complicated topography, diurnal cycle of theearth system shows remarkable regional discrepancy, especially for the precipitation. On one h and, the various underlying surface and terrain respond differentlyto the incident solar radiation, leading to the regionaldifferences of surface thermodynamical forcing. Onthe other h and, they block and damp the air flow differently, forming the regionally varied dynamical forcing. The different thermodynamical and dynamicalforcings are the major factors contributing to the regional variation of diurnal features at the same lati-tude. Diurnal cycle of precipitation reflects the impact of thermodynamical and dynamical processes ofearth system to the hydrological cycle, which is alsoconnected with multi-scale l and -atmosphere, ocean-atmosphere, l and -sea, and mountain-valley interactions, as well as the physical processes involving thephase transformation of vapor, evolutions of cloud and rainfall, and interaction between the atmosphericaerosol and cloud/precipitation water. Meanwhile, thediurnal variations of precipitation can also influencethe hydrological and energy cycles, ecological environment, and the anthropogenic activities, giving feed-back to the earth system. Studying the diurnal cycleof rainfall can be helpful to underst and not only the development and evolution of weather and climate, butalso the mechanisms of rainfall formation.

Because of the importance of the diurnal variation of rainfall, it attracted attention of scientists inas early as the beginning of the 20th century(Hann, 1901). Kincer(1916) pointed out that rainfall at somestations over the central U. S. reaches the maximumduring night while rainfall near coastlines peaks in theafternoon. Using the station data in eastern China, Japan, and Korea, Ramage(1952)found that stations of the coastal regions show early morning rainfallpeaks, and the inl and stations mainly present afternoon rainfall peaks. Chinese meteorologist LÄu(1942) indicated nocturnal rainfall features over the BashanMountain in the mid 20th century. The influences ofmountain-valley and l and -sea effects have also been regonized long ago(Bleeker and Andre, 1951; Neumann, 1951; Kraus, 1963).

Limited by unavailability of hourly precipitationdata, studies of the diurnal rainfall variation were restricted in some local regions and for short periods. With development of the remote sensing technology, exp and ing of metorological observational network, and improvement of numerical models, more and morehigh temporal and spatial resolution observation data and model simulations become available and the studies of diurnal variation of rainfall have been greatlyimproved in the recent 20 years. It is found that thediurnal cycle of rainfall is different over l and and sea and the diurnal signal over l and would spread out overthe adjacent oceans, probably through gravity wavesof varying depths(Dai, 2001; Yang and Slingo, 2001; Sorooshian et al., 2002). The rainfall over l and usuallypeaks in the afternoon with a large amplitude, whichis basically a low-level thermodynamical response tothe radiative heating cycle. This idealized view of thediurnal cycle over l and is modified by local orography and by the initiation, propagation, and decay ofmesoscale convective systems(MCSs), and midnightto early morning rainfall peaks are also found oversome l and area.

Focusing on the two different diurnal peaks overl and, Nesbitt and Zipser(2003)investigated the Tropical Rainfall Measuring Mission(TRMM)data and found that MCSs over all l and regions have rainfallpeaks that occur in the late evening through midnightdue to their longer life cycle, while non-MCSs are featured with a significant rainfall peak in the afternoon. Yang and Smith(2008)pointed out that summer convective rainfall exhibits consistent dominant afternoonpeaks while stratiform rainfall over continents exhibitsa strong late evening peak and a weak afternoon peak. As for the Great Plain of U. S., the eastward propogation of rainfall systems near the Rocky Mountains(Carbone et al., 2002; Tian et al., 2005; Jiang et al., 2006; Carbone and Tuttle, 2008), the mountain-valleyeffects(Wolyn and Mckee, 1994), and the nocturnallow-level jet(Higgins et al., 1997)are all considered asimportant facors leading to the nocturnal rainfall peakthere. For deep oceanic convection, which typicallypeaks in the early morning with a small amplitude, thesituation is much more complicated(Dai, 2001; Yang and Slingo, 2001). There may be several mechanismsresponsible for the behavior of the rainfall diurnal cycle, including a direct radiation-convection interaction(Randall et al., 1991), the cloud versus cloud-free radiation difference in the horizontal(Gray and Jacobson, 1977), and the life cycle of convective systems themselves(Chen and Houze, 1997; Sui et al., 1998).

Surrounded by the Pacific Ocean and TibetanPlateau(TP), the topography and underlying surfaceof China is much complex. China has abundant diurnal rainfall features and significant regional characteristics, making this area an ideal platform for diurnalvariation analysis. Investigating the diurnal variationof rainfall over China can improve our underst and ing of synoptic and climatological system evolutions, as well as the underst and ing of the hydrological and energy cycles and variations of the earth system. Benefiting from the hourly station rain gauge data collected and compiled by the National MeteorologicalInformation Center of the China Meteorological Administration, as well as the springing out of the international high temporal and spatial resolution satellitedata, systematic analyses of the diurnal variation overChina have been carried out in the beginning of the21st century. The present paper will summarize theprogress in research on the diurnal variation of precipitation over contiguous China since the 21th centuryin a comprehensive manner.

The major contents of this paper are dividedinto three sections: the climatological features and regional differences of rainfall diurnal cycle, causes ofregional differences of precipitation diurnal variations, and evaluation of numerical models in simulation ofthe diurnal features of rainfall. At last, a summary and discussion are given in Section 5. 2. Climatic characteristics and regional differences of precipitation diurnal variations2. 1 Diurnal variations of summer precipitation

Using hourly rain gauge data obtained fromthe national climate reference network and nationalweather surface network of China, Yu et al. (2007b)revealed that summer precipitation over contiguousChina has large diurnal variations with considerableregional features. Figure 1 shows the local solar time(LST)of the maximum precipitation and Fig. 2 showsthe diurnal cycle of summer precipitation averagedover five different regions. Over the Sichuan basin and the eastern part of TP, precipitation peaks aroundmidnight, and the diurnal cycle shows a harmonic sinusoidal evolution with the minimum at noon. Precipitation over eastern Sichuan basin and the middle reachesof the Yangtze River(YR)valley shows an early morning maximum around 0600 LST. Both southeasterninl and China and northeastern China have late afternoon(around 1700 LST)maxima of rainfall. The spatial distribution of maximum precipitation over thecentral eastern China is complex. Precipitation in thisregion exhibits a semi-diurnal cycle and is characterized by two comparable peaks(early morning and lateafternoon). The minimum precipitation appears atnoon and midnight.

Fig. 1. Spatial distribution of the diurnal phase of the summer mean hourly precipitation during 1991-2004. Theunit vectors denote the phase clock of the maximum precipitation in local solar time(LST). The red(purple)vectorsrepresent the peaks occurring in 1500-2000(0900-1400)LST, and the blue(green)ones for 2100-0200(0300-0800)LST. [Redrawn based on Yu et al., 2007b]

Figure options

Fig. 2. Diurnal variations of the 1991-2004 mean summer precipitation(mm h-1)over the five regions marked in Fig. 1. (a)Region 1(27°-32°N, 100°-107°E), (b)Region 2(27°-30°N, 108°-113°E), (c)Region 3(23°-26°N, 110°-117°E), (d)Region 4(40°-50°N, 110°-130°E), and (e)Region 5(30°-40°N, 110°-120°E). [Redrawn based on Yu et al., 2007b]

Figure options

Yu et al. (2007b)revealed considerable regionaldifferences in diurnal variations of precipitation, and also pointed out that diurnal phases in different regions may be related. As a typical example, the diurnal phase delays eastward along the YR valley, with amidnight maximum in the upper valley and an earlymorning peak in the middle valley. Comparing Figs. 2a with 2b indicates that the diurnal phase over themiddle reaches of the YR valley lags behind about 6h than the upper valley. The diurnal peak appearsfurther later in the lower valley where many stationspresent the late afternoon maximum. Figure 3 shows atime-longitude cross-section of the meridionally(27°-29°N)averaged hourly precipitation percentage relative to the daily total rainfall. There is an eastwardtransition of the nocturnal maximum over the eastern periphery of TP to the late afternoon peak overthe lower reaches of the YR valley. This phase transition looks like an eastward propagation of convectivesystems initiated over the east periphery of TP. Casestudies, however, indicate that nocturnal heavy rainfall in the upper valley is not always followed by anearly morning heavy rainfall in the middle valley, and an early morning heavy rainfall in the middle valleydoes not always suggest that a nocturnal heavy rainfall has already happened in the upper valley. Themechanism of this phase transition will be further discussed later in this review.

Fig. 3. Time-longitude cross-section(27°-29°N)of me and iurnal variations in hourly precipitation percentage rela-tive to the daily total rainfall amount. [From Yu et al., 2007b]

Figure options

With the rapid development of the meteorologicalsatellite detection technology, various satellite-basedprecipitation products have been widely used. Zhouet al. (2008)analyzed hourly precipitation data fromPrecipitation Estimation from Remotely Sensed Information using Artificial Neural Networks(PERSIANN) and 3-h TRMM 3B42 satellite products to characterize East Asian summer monsoon rainfall. The resultsshow that both satellite products capture the nocturnal peak over the eastern periphery of TP and thelate afternoon peak in the southern and northeasternChina. The satellite products succeed in measuringthe major late afternoon peak but fail in estimatingthe secondary early morning peak in the central eastern China. The quality of satellite products in themiddle YR valley is also questionable with too strong(weak)late afternoon(early morning)rainfall. Exploring why the satellite data overestimate(underesti-mate)the late afternoon(early morning)precipitationin the middle and lower YR valley, Yuan et al. (2012a)compared the rain gauge data with the satellite data. They found that the satellite data with more(less)light rainfall in the afternoon(early morning)may beone of the important reasons for the biases, and thedeviation of the satellite data may be related to thedifferences in the content of ice particles in differentstages of rainfall process.

Restricted by the density of the station network, the diurnal rainfall variation over the western China, especially over the TP, is not fully presented in Fig. 1. There are only a few stations over the southernTP, and they all show coherent nocturnal maximumprecipitation. Fujinami et al. (2005)used hourly geo-stationary meteorological satellite infrared equivalentblack-body temperature data and TRMM precipitation radar(PR)data(product 2A25)to investigatethe diurnal cycle of precipitation over the TP wherethe station rain gauge data are scarce. They foundthat cloud-cover frequency(CCF)for high cloud increased after 1300 LST(0700 UTC)over the mountain ranges along 28. 5° and 30. 2°N, reaching a maximum near 1800 LST(1200 UTC). High values of CCFsubsequently moved to the valley along 29. 3°N. Relatively high CCF persisted over the valley until theearly morning. The rainfall frequency in TRMM PRdata shows a similar evolution with high cloud, reaching a maximum around the late afternoon(midnight)over the mountain range(valley). Since most stationsover the TP are located in valleys, the precipitationusing rain gauge data shown in Fig. 1 has a nocturnalmaximum. The diurnal variation differences relatedto topography will be thoroughly discussed later. 2. 2 Seasonal and sub-seasonal variations ofrainfall diurnal cycle

Li et al. (2008b)studied the seasonal variation ofthe diurnal cycle of rainfall in the southern contiguousChina. They pointed out the differences of precipitation diurnal phase between warm(May-September) and cold seasons(November-March). As illustrated inFig. 1, the diurnal peaks of summer rainfall over thesouthern contiguous China show distinct differencesbetween the west and east: the southwestern Chinatends to reach the rainfall maximum in either themidnight or early morning, while precipitation in thesoutheastern contiguous China has a coherent peak inthe late afternoon. In cold seasons, the diurnal phasedifferences between these two regions are not obvious. As shown in Fig. 4a, the entire southern China showscoherent midnight or early morning diurnal rainfallmaxima during the cold seasons. In contrast with theweak seasonal variation of the diurnal phases of pre-cipitation in the southwestern China, the rainfall peakin the southeast shifts sharply from warm seasons tocold seasons. The rainfall in most stations(72. 2%)ofthe southeastern China reaches the maximum in themidnight and early morning(0000-0800 LST)in thecold seasons. Only at four stations does the maximumrainfall appear in the afternoon and evening(1300-2100 LST)|two of them are located in mountain areas and the other two are located in southern coastalregions. Figures 4b and 4c show seasonal changes ofthe diurnal cycle in the southwestern and southeastern China. In cold seasons, diurnal phases in bothregions exhibit a single early morning peak. Duringthe warm seasons, the maximum rainfall in the south-western China still occurs in the early morning, and a secondary afternoon(1500 LST)peak can only befound in August. The maximum hourly rainfall insoutheastern China, in contrast, occurs at 1600-1700LST in all warm seasons.

Fig. 4. (a)Spatial distribution of the diurnal phase of the 1991-2004 mean diurnal cycle of hourly precipitation forcold season(November-March). The unit vectors denote the phase clock of the maximum precipitation in LST. Twodistinct regions are labeled: the southwestern and southeastern China. The diurnal cycle of precipitation, st and ardizedin each month, for the(b)southwestern and (c)southeastern China. [Redrawn based on Li et al., 2008b]

Figure options

Most precipitation in the monsoon region over theeastern China occurs in warm seasons. Influenced bythe monsoon circulation and the meridional movementof monsoon rain belt, the diurnal phase of precipitation in warm seasons has a significant sub-seasonalvariation. Based on the TRMM Microwave Imager(TMI)data, Chen et al. (2009)grouped the precipitation data in warm seasons into presummer(May-June) and midsummer(July-August). They pointedout that the diurnal amplitude increases in presummerthan before, and the strong morning rainfall dominatesmost areas of southeastern China; during midsummer, the afternoon peak rainfall becomes a notable featureover the southern China, while the morning rainfall issignificantly reduced. Yuan et al. (2010)analyzed thesub-seasonal characteristics of the diurnal variation ofthe summer monsoon rainfall over the central easternChina using hourly station rain gauge data. As themonsoon rain belt undergoes stepwise movement, fourperiods(mid and late June, early and mid July, lateJuly to mid August, and late August to early September)have been selected to represent the major st and ing stages of the monsoon rain belt over the centraleastern China.

Figure 5 shows the diurnal peaks of rainfall forthe four periods. In general, the rainfall in warm seasons over the central eastern China mainly peaks inthe early morning and late afternoon, which is coherent to the previously described characteristic of double peaks. By comparing the four periods, the earlymorning diurnal peaks exhibit b and -like spatial distributions, and the early morning rainfall shows thesub-seasonal movement coinciding with that of themonsoon rain belt. During mid and late June, theearly morning peak b and mainly extends along thearea around 30°N. In the early and mid July, the earlymorning peak moves to the Huai River valley. Duringlate July to mid August, the early morning rainfallpeaks shift northward to the northern China, coinciding with the migration of the monsoon rain belt. Inlate August to early September, early morning rainfallpeaks withdraw to the south and dominate the HuaiRiver valley again. With the early morning peak b and moving northward and southward, the stations to thesouth and north of the morning peak b and mostlyshow late afternoon rainfall peaks. Four regions of theearly morning peak during four periods are labeled inFig. 5. According to whether the summer monsoonrain belt is in these four regions or not, they partitioned the whole summer into active and break monsoon periods and found that the rainfall at differentlatitudes of the central eastern China exhibits similardiurnal features. During the active period, monsoonrainfall is dominated by the early morning peak. Incontrast, during the break period, monsoon rainfall isdominated by the late afternoon diurnal peak.

Fig. 5. Spatial distributions of the diurnal phase and amplitude(% of the daily mean)of rainfall diurnal cycle averagedin(a)mid and late June, (b)early and mid July, (c)late July to mid August, and (d)late August to early September. The shading represents the amplitude, and unit vectors denote the phase clock in LST of the maximum precipitation(see phase clock and colors). The green(blue)vectors represent the diurnal peaks occurring between 0200 and 0800(2000 and 0200)LST, and the red(black)vectors for the diurnal peaks occurring between 1400 and 2000(0800 and 1400)LST. Four distinct regions are also labeled. [From Yuan et al., 2010]

Figure options

Xu and Zipser(2011)also divided the warm seasons into 3 periods("pre-Meiyu" from 1 April to11 May, "Meiyu" from 15 May to 25 June, and "midsummer" from 1 July to 10 August)to investigate the seasonal variability of the precipitation diurnal cycles and relevant systems. They found thatover the eastern China, deep convection usually hasan afternoon peak during the Meiyu, whereas the rainfall maximum is in the early morning. In midsummer, most of East Asia is dominated by afternoon precipitation and convection, with the northern part of easternChina as an exception, where early morning rainfallprevails in spite of most deep convection and lightningoccurring in the afternoon. Most of the midnight and early morning rainfall is contributed by MCS. Alsoconcerned about the sub-seasonal evolution of the diurnal variation of rainfall in East Asia, Geng and Yamada(2007)designed a front-relative coordinate system, with the 0°-latitude centered at where the front is(longitude is still the actual longitude). Based on thiscoordinate system, they analyzed the precipitation diurnal variations of Meiyu front and the variations overthe southern and northern sides of the front. The results show that the reflectivity observed by TRMMPR in the frontal zone has a remarkable diurnal cycle that peaks in the early morning. Meanwhile, thereflectivity to both the south and north of the frontalzone has a diurnal peak in the late morning and afternoon.

The seasonal variability of rainfall diurnal cycleis reflected not only in the variation of diurnal phasesbut also in the relation of precipitation diurnal phasesamong different regions. Bao et al. (2011)partitioned warm season into three different month-longperiods: the pre-Meiyu period(15 May-15 June), theMeiyu period(15 June-15 July), and the post-Meiyuperiod(15 July-15 August). They used the high-resolution NOAA/Climate Prediction Center morphing technique(CMORPH)precipitation data to explore the characteristic of eastward propagation of diurnal phase during different rainy periods. The resultsshow that the diurnal peak, originating from the eastern edge of TP during the pre-Meiyu period, propagates eastward at an average speed of about 15 ms^-1. It reaches the eastern coast of China on the 3rdday(approximately 45 h)covering a distance of about2000 km. The average eastward speed of this diurnalprecipitation peak is approximately 13 m s^-1 duringthe Meiyu period, which is clearly traceable only tothe western edge of the plains(approximately 114°E)in about 32 h. The post-Meiyu period features theslowest propagation speed of about 9 m s^-1, whichis mostly confined to the west of 113°E and barelyreaches the plains. Xu and Zipser(2011)obtainedsimilar results based on the analysis of TRMM data:diurnal cycles of rainfall present apparent phase propagation eastward from the eastern TP, and the phasepropagation is most evident during the pre-Meiyu and Meiyu seasons. However, it weakens with the north-ward progress of the East Asian monsoon and ceasesin midsummer. 3. Causes of regional differences of precipitation diurnal variations3. 1 Regional differences of diurnal variationsof different rainfall events

The mainl and of China not only occupies thelargest continental area in the East Asian monsoonregion, but also is characterized by complex l and -seacontrast and irregular mountain-valley distributions. Therefore, the diurnal variation of rainfall over contiguous China exhibits unique climatic characteristicswith evident regional differences, which is closely related to the evolution of different rainfall events. According to duration time, rainfall events be classifiedinto long- and short-duration events; while accordingto associated cloud structure, they can be divided intostratiform and convective ones. 3. 1. 1 Long-duration vs. short-duration rainfall events

Yu et al. (2007a)classified the rainfall eventsaccording to their duration time: those that last forless than or equal to 3 h are defined as short-durationevents, and those that last for longer than 6 h as long-duration events. It is found that over 50% of rainfallevents have a duration of 1-3 h, mostly in the north-eastern China and southern inl and of China, whilelong-duration rainfall events occur most frequently(over 20%) and contribute more than 60% to the total rainfall in the central eastern China. Along theYR valley, over 66% of the warm season rainfall comesfrom the contribution of long-duration events(see Fig. 1 in Yu et al., 2007a). The short- and long-durationrainfall events present different diurnal variations inwarm seasons. The double peaks of rainfall in thecentral eastern China are closely related to the rain-fall duration: the maxima of short-duration rainfallevents usually appear in the late afternoon, while thepeaks of long-duration events prefer to happen in theevening and early morning(Fig. 6). In addition, the diurnal delay of rainfall peaks along the YR valley is mainly contributed by the long-duration rainfallevents(Chen et al., 2010). The short-duration rain-fall events are dominated by afternoon peaks due todevelopment of the atmospheric boundary layer or theincreasing sensible and latent heat flux from surfaceinduced by the solar heating, thus the diurnal phaseof these events shows no evident regional differences. The peak hours of long-duration rainfall events exhibitzonal and meridional differences: the nocturnal long-duration rainfall presents an eastward-delayed diurnalphase of rainfall down the YR valley; in the easternChina, the long-duration rainfall events reach the diurnal maxima before 0600 LST in the north of theHuai River, while they peak around 0800 LST in thelower reaches of the YR valley and the south of theHuai River(Yuan et al., 2014).

Fig. 6. Spatial distributions of the diurnal phase of sum-mer mean(a)long- and (b)short-duration rainfall events. The unit vectors denote the phase clock in LST of themaximum precipitation. [From Yu et al., 2007a]

Figure options

The diurnal cycles of rainfall events also show evident seasonal and inter-decadal variations. Diurnalchanges of short- and long-duration rainfall events ineach month averaged over the southwestern and south-eastern China are presented in Fig. 7. For rainfallevents lasting for 1-3 h, the diurnal phases in both regions show a dominant peak in the afternoon in warmseasons. The afternoon peak in southeastern China exists from February to October. During February and March, the diurnal precipitation peaks at 1400 LSTin southeastern China. A morning peak ranging from0600 to 0800 LST develops during November-January. However, in the southwestern China, the afternoonmaximum only occurs from May to September. Forrainfall events lasting more than 6 h, the southwesternChina experiences a negligible annual change in diurnal phases. The rainfall diurnal cycle is dominatedby an early morning peak throughout the year. Incontrast to the coherent diurnal phases in each seasonin the southwestern China, a prominent seasonal variation of long-duration precipitation is seen over thesoutheastern China. From January to June, the diurnal peak is postponed gradually from 0200 to 0900LST. From July to September, the maximum rain-fall amount appears in the late afternoon. From October to December, the diurnal peak appears in theearly morning(Li et al., 2008b). By analyzing thedecadal change in the duration-related characteristicsof late summer(July-August)precipitation over theeastern China during 1966-2005, Li et al. (2011)found that the duration-related rainfall structure experienced significant interdecadal changes accompanying the "southern flooding and northern drought" pattern of rainfall amount over the eastern China in recentdecades. Specifically, in North China, the frequencyof short-duration rainfall events decreased and theirintensity increased, and the decadal decreases of rain-fall amount over North China were largely contributedby long-duration rainfall events, especially those occurring between midnight and morning. Analysesbased on hourly records from Beijing station also confirmed that the total rainfall amount of short-durationevents has increased significantly, while the total rain-fall amount of long-duration events has decreased(Li et al., 2008a). In the mid-lower reaches of YR valley, both the frequency and amount of long-durationprecipitation have significantly increased. Yuan et al. (2013a)revealed that over the south of YR, as wellas the area between the Yangtze and Yellow rivers, the percentages of morning rainfall to total rainfall interms of amount, frequency, and intensity, all exhibited increasing interdecadal trends, while over NorthChina, decreasing trends were found. Since the long-duration rainfall is closely related to the large-scalemonsoon circulation and monsoon rain belt, the interdecadal variation of these events is closely associatedwith the strong tropospheric cooling and the weakening of summer monsoon over East Asia(Yu et al., 2004b; Yu and Zhou, 2007; Zhou et al., 2009).

Fig. 7. Seasonal variations of the diurnal cycle of(a, b)short- and (c, d)long-duration precipitation, st and ardized ineach month for(a, c)southwestern and (b, d)southeastern China. The x-axis signifies the change of local solar time and the y-axis signifies the annual cycle. Dashed contours denote negative values. The regions for average are markedin Fig. 4a. [From Li et al., 2008b]

Figure options

The total rainfall is constituted of the stratiformprecipitation associated with relative stable stratification, and the convective precipitation related to vigorous overturning(Houze, 1993, 1997; Tao et al., 1993). Base on TRMM 2A25 products measured by TRMMPR in summer, Yu et al. (2010)pointed out thatthe convective and stratiform rainfall contribute similar proportions to the total rainfall averaged over thesouthern China, but the spatial distribution is inhomogeneous: the stratiform(convective)rainfall dominates the southwestern(southeastern)China, whilethe propotions of the two kinds of rainfall are approximately equal over the region between the Yangtze and Huai rivers. The convective and stratiform rainfallsshow distinct diurnal variations. The diurnal peak ofthe stratiform rain rate presents a sharp contrast between the southwestern and southeastern China. Thestratiform rainfall peak occurs during the late nightover 52. 5% of the southwestern China, while the stratiform precipitation peaks in the late afternoon over63. 8% of the southeastern China(Fig. 8a). However, the phase contrast between the southwestern and southeastern China disappears in convective rainfall(Fig. 8b). Over 76. 2% of the southern contiguousChina, convective rainfall shows a late afternoon peak. Nevertheless, over the eastern edge of TP, the convective rain rate peaks around the late night. The timewhen the maximum hourly rainfall occurs also dependson the rainfall duration. By combining the rain gaugedata and TRMM 2A25 products, the stratiform and convective rainfalls were classified according to theirduration. The results showed that stratiform rainfallusually lasts for a longer time, while convective rain-fall mainly presents short-duration features. For long-duration rainfall events, the diurnal peak of stratiformprecipitation occurs in the late night over 62. 3% ofthe southwestern China. The convective long-durationprecipitation exhibits a maximum hourly rainfall inthe late night over 54. 9% of the southwestern China. Over the southeastern coastal area, the late afternoonpeak is predominant for either stratiform or convective long-duration precipitation(Figs. 8c and 8d). Forshort-duration rainfall events, stratiform precipitationover 55. 1% of the southern China peaks in the late afternoon. The late afternoon peak of convective precipitation controls 75. 7% of the southern China(Figs. 8e and 8f). Over the eastern periphery of TP, both convective and stratiform rainfalls peak in the late night.

Fig. 8. Spatial distributions of the diurnal phase(in LST)of summertime mean(a)stratiform rain rate, (b)convectiverain rate, (c)stratiform long-duration rainfall events, (d)convective long-duration rainfall events, (e)stratiform short-duration rainfall events, and (f)convective short-duration rainfall events, detected by TRMM PR during 1998-2006. Two distinct regions are labeled in(a) and (b). [Redrawn based on Yu et al., 2010]

Figure options

In addition to the rain rate, the profiles of convective and stratiform rainfall also exhibit evident diurnalvariation differences(Liu and Fu, 2010). The diurnalfeatures of regional averaged rainfall profiles were assessed by Yu et al. (2010). Both the rain top and nearsurface rain rates of stratiform precipitation peak atmidnight in the southwestern China(Fig. 9a). In thesoutheastern China(Fig. 9b), the rain top of strati-form precipitation peaks in the evening, but the nearsurface rain rate attains the maximum in the afternoon. Compared with the stratiform rain rate profile, the convective precipitation profile shows muchstronger diurnal variations(Figs. 9c and 9d). Theconvective rain top and near surface rain rates in boththe southwestern and southeastern China peak in theafternoon. During the 4 h between the noon and afternoon, the rain top shows the maximum rise. Thisuniform quick strengthening of convection might becaused by the favorable thermal condition in the afternoon. Over the southern contiguous China, the minimum stratiform precipitation occurs in the noon and the convective precipitation reaches the lowest level inthe morning.

Fig. 9. Diurnal variations of the(a, b)stratiform and (c, d)convective precipitation profiles averaged over the(a, c)southwestern and (b, d)southeastern China. The regions for average are marked in Fig. 8a. [From Yu et al., 2010]

Figure options

To sum up, both the short- and long-durationrainfalls, and convective and stratiform rainfallspresent distinct regional characteristics, which reflectthe different contributions of dynamical and thermalforcing to the rainfall diurnal cycle in varied areas. The afternoon short-duration and convective rainfallsare mainly forced by local thermal forcing, while thenocturnal to early morning long-duration and stratiform rainfalls are in close correlation with large-scalecirculations. 3. 2 Influences of mountain-valley breeze and l and -sea breeze

In the central eastern China, diurnally forcedl and -sea and terrain-slope differential heating initiatevaried local circulations(Yu et al., 2009), which support organized cloud clusters that are characterized bya pronounced diurnal cycle(Yin et al., 2011; Wang et al., 2012; Zhuo et al., 2013). He and Zhang(2010)argued that the mountain-plain solenoid(MPS)circulation plays an important role in the regional differences of rainfall diurnal variations in mountain and plain areas over North China in warm seasons. Theyfound that the averaged peak in local precipitation begins early in the afternoon near the top of the mountain ranges and propagates down slope and south-eastward. The peak reaches the central North ChinaPlains around midnight to early morning hours, resulting in a broad area of nocturnal precipitation maximaover the plains. The diurnal precipitation peak(minimum)is closely collocated with the upward(down-ward)branch of an MPS circulation. Both the MPScirculation and the low-level southwesterly nocturnaljet are likely responsible for the nighttime precipitation maxima over the plains.

Based on hourly intensive surface observationsduring June-July 2010, Zhao et al. (2012)also pointedout that the local thermal forcing contributed largelyto the rainfall diurnal variation over regions with different orographic features along the middle reaches ofthe YR valley, as the spatial distribution of surfacetemperature diurnal changes is influenced by the to-pography. The temperature rising(dropping)is largerin the mountainous areas than in the plains duringdaytime(nighttime), which results in the near surfacethermal flow driven by the diurnal change of temperature gradient due to the topographic difference. Thus, the convergence(divergence)flows occur in the mountainous(plain)areas in the daytime, and vice versa inthe nighttime. The decrease of stability and increaseof convective available potential energy(CAPE)in thelower atmosphere are beneficial to the afternoon thermal convection initiation, while the lowering of liftingcondensation level and increase of relative humidity and precipitable water vapor are favorable for the development of early morning convection.

The coexistence of late afternoon and nocturnalrainfall peaks over the TP are proposed to be highlyrelated to the mountain-valley breeze caused by thecomplex regional topographic features(Fujinami et al., 2005; Singh and Nakamura, 2009). Analyses of recordsfrom a 3-yr intensified observational experiment at8 stations along the hill side of Seqilashan over thesoutheastern TP also demonstrated the relationshipbetween regional differences of rainfall diurnal variation and the local thermal forcing caused by topography(Chen et al., 2012b). The prevailing nocturnalrainfall peak in the observations at routine stations canbe largely attributed to the relatively lower location ofthe stations, which are mostly situated in valleys. Therecords at Seqilashan stations on hillsides revealed anevident afternoon peak of warm season rainfall(Fig. 4in Chen et al., 2012b). The different diurnal phases between valley and hillside stations are closely related tothe orographically induced regional circulations overthe TP. The topography-induced mountain-valley circulation is consistent with different diurnal phases atvalley and hillside stations. The wind blows towardthe high terrain in the daytime, which favors the formation and development of convection over hillsides. In nocturnal hours, the wind blows into the valley, up-lifting the air and thus favoring the convection in thevalley(Fig. 5 in Chen et al., 2012b).

In addition to the influence of local topography and l and -sea distribution, the large-scale "mountain-valley" and "l and -sea" breezes could also contributeto the rainfall diurnal features in the subtropical regions of China(Yuan et al., 2012b). The TP, with amean elevation over 4000 m, is a strong heat sourcein summer and sink in winter, functioning as an important modulator of the Asian climate(Ye and Gao, 1979; Kuo and Qian, 1981; Wu et al., 2006; Wu et al., 2007). The plateau can be regarded as a "mountain"relative to the "valley", such as the Sichuan basin and the East China plains. On the diurnal timescale, thethermal effect of TP may also be a great factor to thediurnal feature over East Asia. In the afternoon toearly evening, a warm diurnal temperature trough at700-500 hPa can be found over the TP. Corresponding to the thermal conditions, strong upward motionsoccur over the plateau(Fig. 10 in Yuan et al., 2012b). Averaged over the TP, the maximum pressure vertical velocity exceeds 0. 4 Pa s^-1 near the surface ofthe plateau. Meanwhile, the maximum of moist staticenergy temperature in the lower troposphere is quiteclose to the minimum of saturated moist static energytemperature at 1400 Beijing Time(BT)(Fig. 11 in Yuan et al., 2012b), which means that the thermal and moist condition is conducive to the occurrence of convection in the afternoon over the plateau. Influencedby the TP, deep continental stratus cloud is found overthe Southwest China plain(Yu et al., 2004a), whichhinders the solar radiation from reaching the ground(Li et al., 2008b). Combined with the eastward wind inthe middle and upper troposphere, an apparent warm and moist advection locates over the upper level ofthe southwestern plain(Chen et al., 2010). All thesefactors would suppress the development of static instability and the afternoon local thermal convectionover the southwestern plain. During night, the zonal and vertical winds over the TP and the southwesternplain reverse. The plateau is controlled by subsidingair throughout the troposphere. The anomalous airflow toward the TP causes low-level convergence and updrafts over the plateau's eastern slope at night. Accompanied by the long wave radiative cooling at thecloud top(Lin et al., 2000) and the weak cold advection that has decreased the stability(Chen et al., 2010), the rainfall over the southwestern plain reachesthe maximum during the midnight and early morning.

Fig. 10. (a)Composite JJA mean of 850-hPa wind anomalies(vectors; m s-1)at 2000 BT in long-duration rainfalldays for 1991-2004. The shading denotes the normalized long-duration rainfall amount accumulated during 2000-0200BT. (b)Composite anomalous zonal circulation averaged between 27° and 33°N at 2000 BT. The gray shading denotesthe meridionally averaged vertical velocity(10-2 Pa s-1), and the black denotes the topography. Panels(c, d)are thesame as in(a, b), but for 0800 BT. [Redrawn based on Chen et al., 2010]

Figure options

Fig. 11. Diurnal variations of normalized(by the daily mean)(a, d)total, (b, e)convective, and (c, f)stratiform rainfallin observations(thick solid lines), and fine(dashed lines) and coarse(thin solid lines)resolution CAM5 outputs averagedin the western(28°-35°N, 10°-108°E; a, b, c) and eastern(28°-35°N, 112°-120°E; d, e, f)plains. The percentages ofconvective(stratiform)rainfall to total rainfall from the coarse and fine resolution CAM5 runs and the observations aremarked on the top-left sides of panels(b, e)((c, f)). [Redrawn based on Yuan et al., 2013b]

Figure options

The central and eastern plains can be regarded asthe "valley" with respect to the TP, as well as the"l and " relative to the East China Sea. Huang W. R. et al. (2010)proposed that the local l and -sea breezecouples with the global-scale diurnal atmospheric pressure tide, producing a planetary-scale l and -sea breezewith a spatial scale of 1000 km, which may be one ofthe possible mechanisms for the diurnal variation ofsummer rainfall in the subtropical East Asia. Becauseof the influence of both the large-scale "sea breeze"in the afternoon and "mountain wind" at night, thepercentages of late afternoon and early morning rain-fall are fairly close to each other over the central and eastern plains. The late afternoon rainfall graduallyincreases from the eastern edge of the southwesternplain to the coastlines. This may be a result of thecompetition between the large-scale mountain-valleycirculation and the l and -sea circulation. In addition, positive CAPE exists in the afternoon over l and, and upward vertical motion occurs around 1400 BT overthe eastern plain. At night, negative pressure verticalvelocity(i. e., updraft)occurs at 0200 BT throughoutthe boundary layer to the mid-troposphere, reflectingthe role of the eastern plain as a valley. Except forvertical velocity, the 850-hPa southwesterly flow overthe eastern plain strengthens during night. The south-westerly flow contributes to the formation of the low-level jet and transports the warm tropical water vaporto the eastern plain, providing a favorable environmentto the rainfall systems over this area. 3. 3 Influences of the lower troposphere circulationThe eastward transition of the nocturnal rainfallmaximum over the eastern periphery of TP to the lateafternoon peak over the lower reaches of the YR valleyis a salient characteristic of the rainfall diurnal variation over contiguous China. Bao et al. (2011)indicated that the local peak of the summer diurnal precipitation usually begins in the mid to late afternoonon the eastern edge of TP, and subsequently propagates eastward and down slope. This is strongly associated with several thermally driven regional MPSsdue to the differential heating between the plateaus, the highl and s, the plains, and the ocean. They alsoshow that there is a substantial difference in the propagation speed and eastward extent of the peak phaseof the dominant diurnal precipitation cycle originatedfrom the TP. This diurnal peak has a faster(slower)eastward propagation speed and a more(less)coherent propagation duration, and thus covers the longest(shortest)distance to the east during the pre-Meiyu(post-Meiyu)period than that during the Meiyu period. The differences in the mean midlatitude westerly flow and in the position and strength of the western Pacific subtropical high during different periodsare the key factors in causing the differences in thepropagation speed and eastward extent of this dominant diurnal precipitation cycle. Based on a WRFmodel simulation, Huang H. L. et al. (2010)found thatthe eastern TP facilitated convective development inthe afternoon, and the solenoidal circulation betweenthis region and its leeside lowl and s(near the Sichuanbasin)also contributed to the longevity and fartherdownstream propagation of the rainfall episodes under prevailing westerly winds.

Nevertheless, nocturnal rainfall in the upperreaches of the YR valley is not always followed bythe early morning rainfall in the middle valley, and the early morning rainfall in the middle valley doesnot always suggest a preceding upstream night rain-fall(Yu et al., 2007b). Therefore, the downstreamphase transition is not always caused by the eastwardpropagation of convective systems. According to Yuet al. (2007a), long-duration rainfall events dominatethe summer rainfall along the YR. Thus, the local effects may not adequately explain the nocturnal long-duration rain that is closely related to the large-scaleforcing(Chen, 1983). The intimately coupled monsoon system components, especially the southwesterlylow-level jet, are crucial factors for maintaining therainfall along the YR(Chen and Li, 1995; Wang et al., 2000; Ding et al., 2001). It has been demonstratedthat the long-duration rainfall events are closely associated with the enhanced southwesterly(Chen et al., 2010). The combination of the southwest vortex and the shear line, as well as the low-level convergence ofthe warm tropical water vapor with the cold subtropical water vapor, provides a favorable environment formaintaining the rainfall systems along the YR valley.

The lower tropospheric wind shows notable diurnal variations(Lindzen, 1967; Wallace and Tadd, 1974; Krishnamurti and Kishtawal, 2000), whichgreatly influence the rainfall diurnal variation. Theanomalous low-level wind vectors(with the daily meanremoved)generally rotate clockwise during the day. The diurnal clockwise rotation of anomalous low-level wind corresponds well to the preferred nocturnal occurrence of the long-duration rain and itseastward phase transition(Chen et al., 2010). Thelong-duration rainfall events tend to start during thenight and to peak after several hours of development. The eastward-delayed initiation of the nocturnal long-duration rainfall events is thought to be due to the diurnal clockwise rotation of the low tropospheric circulation, especially the accelerated nocturnal southwesterly. In the early evening(2000 BT), the anomalouseasterly flow toward the TP causes low-level convergence over the plateau's eastern slope, inducing theformation of rainfall in the upper YR valley, corresponding to the frequent initiation of long-durationrainfall events around this time. In the midnight(0200BT; Figs. 10a and 10b), the anomalous wind sequentially rotates clockwise to a southerly flow at midnight and accelerates the meridional wind in the middle valley, resulting in the initiation of rainfall between 2300 and 0300 LST(Fig. 8 in Chen et al., 2010). In theearly morning(0800 BT; Figs. 10c and 10d), theaccelerated southwesterlies in southern China, whencombined with decelerated winds in the north of YR, cause strong convergence along the YR valley and contribute to the early morning rainfall in the lower valley. The anomalous convergence zone and upward motionalso show sub-seasonal movement coinciding with themonsoon rain belt, corresponding to the early morning peak of rainfall along the monsoon rain belt(Yuan et al., 2010; Chen et al., 2013). Meanwhile, the different occurrence time of anomalous low-level convergence and upward motion between the southern and northern parts of the Huai River is one of the important reasons that cause the different peak time of long-duration rainfall events between the two regions(Yuan et al., 2014). Based on TRMM products, Chen et al. (2012)also proved the relationship between wind and rainfall on the diurnal timescale.

Moreover, Huang and Chan(2011)indicated thatthe thermal advection caused by diurnal variationof low-level wind can explain the early morning secondary peak of rainfall in South China. Diagnosesof the atmospheric thermodynamic conditions indicatethat late night vertical differential thermal advection and semi-diurnal variation in l and -sea differential radiative heating/cooling are the major reasons for reduction in stability in the early morning and, in turn, facilitate the formation of an early morning maximumin rainfall. Analysis of the water vapor budget further suggests that the early morning maximum rainfall over Southeast China is mainly maintained by thesemi-diurnal harmonic of water vapor flux transportedfrom the South China Sea. 4. Model simulated diurnal rainfall variationsover China

Analyzing the rainfall diurnal variation is not onlyan important way to investigate the formation and evolution of rainfall processes, but also a good metric for assessing the skill of numerical models and underst and ing the nature of model errors(Slingo et al., 1987; Randall et al., 1991; Lin et al., 2000). In thissection, the ability of numerical models in replicatingthe basic rainfall diurnal features over China, and thephysical processes that have mainly contributed to thesimulated biases, were discussed. 4. 1 Simulation results

For the regional models, Dai(2010)firstly comprehensively assessed the regional weather modelAREM in simulating the diurnal cycle of summer rain-fall over China. The results showed that AREM canreproduce the late afternoon peak over the southeastern coastal regions and northeastern China and thenocturnal rainfall maximum in the Sichuan basin. Thesimulated and observed diurnal series are positivelycorrelated over most parts of China and the correlationcoefficient exceeds 0. 6 over the southern and north-eastern China. However, the secondary peak in theearly morning over the middle and lower reaches ofthe YR valley is underestimated and the area of nocturnal rainfall peaks in the simulation is smaller and locates too westward than the observed. Focusing onthe Sichuan basin, Shen and Zhang(2011)pointed outthat the RegCM3 can reproduce the nocturnal peak inthe Sichuan basin and it is mainly contributed by theconvective rainfall, while the amplitude of stratiformrainfall is relatively weak.

For the global models, Yuan(2013)analyzed theAtmospheric Intercomparison Project(AMIP)simulations of the summertime diurnal cycle of precipitation over subtropical China by using outputs fromthe Intergovernmental Panel on Climate Change FifthAssessment Report(IPCC AR5)models. All thesesix global climate models(GCMs)captured the earlymorning peak of total rainfall over the East ChinaSea, but none of these models were able to reproduce the broad geographical patterns of diurnal rain-fall variations over l and areas of subtropical China(Fig. 5 in Yuan, 2013). HadGEM2-A, INMCM4, and MRI-AGCM3-2H simulated the near local noon diurnal peaks over most regions to the east of TP and missed the nocturnal rainfall in the Sichuan basin and middle-lower reaches of the YR valley. The precipitation simulated by BCC-CSM1-1 and FGOALS-g2mainly occurred during midnight over l and s, whilethey cannot reproduce the late afternoon rainfall maximum observed over most l and areas. Although thepeak in the Sichuan basin simulated by MRI-CGCM3appeared near noon, the ratio of nocturnal rainfall tototal rainfall is higher in the Sichuan basin than ineastern China, which is closer to the observation(Fig. 5 in Yuan, 2013). Wang and Zhang(2009)examinedthe simulated diurnal cycle of summer precipitation byMIROC. They found that the amplitude in the modelsimulation was relatively weak, the nocturnal rainfallpeaks were situated far too west, and the early morning peak over the middle and lower reaches of the YRvalley was poorly represented. The afternoon peakover the southern China was delayed to 2100 BT, compared with the observation at 1700 BT.

To summarize, simulations of diurnal cycle of precipitation are greatly model dependent. Some modelsfail to adequately capture the regional differences and cannot simultaneously represent the afternoon peakover southern China and the midnight peak in theSichuan basin. Some models can partly replicate theoverall pattern, but the simulated nocturnal rainfallregions were usually too small and located too westwhen compared with the observation. Meanwhile, thesimulated rainfall in local afternoon tended to rain several hours earlier or later than the observed. 4. 2 Factors influencing the simulation4. 2. 1 Horizontal resolution

The diurnal cycle of precipitation is affected bythe thermodynamical and dynamical factors. Increasing the horizontal resolution of models can better represent the topography and l and -sea contrast, whichmay improve the modeled diurnal cycle. Satoh and Kitao(2013)compared the diurnal cycles of rainfalldescribed in NICAM-11. 2 and -5. 6-km(horizontal resolution)over southern China. They pointed out thatthe precipitation peak occurred at around 1700 LSTin NICAM-5. 6-km, which is similar to the time of peakin TRMM 3B42; conversely, the precipitation peakof NICAM-11. 2-km lagged these peaks by about 2 h. Further analyses showed that moist static energy inthe finer resolution model is produced earlier than thatin the coarse resolution one, which may contribute tothe bias mentioned above.

However, large errors in phase and amplitude ofthe rainfall diurnal cycle may still exist in increasedhorizontal resolution simulations. Comparing the simulations of AREM at different resolutions, Dai(2010)found that when the resolution increased from 75 to 37km, AREM showed a definitely improved diurnal cycle. When the resolution further increased to 15 km, the description of diurnal variations became poorerinstead, as an afternoon rainfall peak appeared in theSichuan basin. Besides, the diurnal cycle of convective and stratiform rainfall produced in fine and coarse resolution CAM5 barely showed any differences(Yuan et al., 2013b). The stratiform rainfall in the two simulations reached the maximum in the early morning, similar to the satellite data(Figs. 11c and 11f). Theconvective rainfall both occurred at 1400 LST, earlierthan the observed, and the sencondary midnight peakin the Sichuan basin was missed(Figs. 11b and 11e). The amplitude of the fine resolution version intensified(Figs. 11a and 11d) and the ratio of stratiformrainfall increased, when compared with the coarse resolution one. Increasing the horizontal resolution ofnumerical models has not really sovled the problemin simulating the diurnal cycle of rainfall. It may beattributed to the fact that the improved circulationcan only have effect to the precipitation diurnal cyclethrough physical parametrization, which is at presentstill different from the real rainfall process. As a result, the crutial factors affecting the simulated diurnalvariations of rainfall still exist in the rainfall relatedphysical parametrization. 4. 2. 2 Physical processes

One of the feasible ways to determine the influence of physical parametrization on the model produced diurnal variation of rainfall is to analyze the diurnal variation of rainfall in the reanalysis data. Daiet al. (2011)calculated the ratio of rainfall in fourperiods to the total rainfall in four reanalysis datasets(NCEP1, NCEP2, ERA40, and JRA25) and the ratios in reanalysis data exhibit large deficiencies whencompared with station observations(Fig. 12). Inrain gauge data, the rainfall over eastern China(theSichuan basin and TP)mainly occurs during 1400-2000(2000-0800)BT, while the similar case is during0200-0800(0800-1400)BT in NCEP1. Precipitationin NCEP2 and EAR40 mainly appears during 0800-1400 BT, which is the period with the least rainfallin reality. The diurnal cycle of rainfall in JRA25 isrelatively close to the observation, but the afternoonrainfall is stronger and the nocturnal rainfall is weaker. The results revealed that the errors caused by parame-terization were still significant, even when models wereforced by the "perfect" large-scale circulation.

Fig. 12. Spatial distributions of the percentage of summer rainfall occurring during 0200-0800(1st column), 0800-1400(2nd column), 1400-2000(3rd column), and 2000-0200 BT(4th column)to total rainfall revealed by the station data(1st row), NCEP1(2nd row), NCEP2(3rd row), ERA40(4th row), and JRA25(5th row)during 1991-2002. [From Dai et al., 2011]

Figure options

The simulation of diurnal cycle of rainfall involvesthe cumulus convection, planetary boundary layer, l and surface, radiative transfer processes and manyothers. To identify the contribution of each process tothe modeled rainfall diurnal cycle, Huang et al. (2008)studied results of a regional climate model and foundthat the l and surface and radiative transfer schemeshave less effects on the summer precipitation diurnalcycle; whereas the simulation of diurnal rainfall variation is very sensitive to the choice of convective cumulus scheme. The Grell scheme realistically capturesthe diurnal variation of summer rainfall over the TP and the middle YR Valley. The Kuo and Anthes-Kuoscheme can well reproduce that over the northeastern and southern China. All of these three schemes cannot simulate the two peaks over the Jiang-Huai region and North China. Analysis of the simulations of theNCEP Regional Spectral Model and the WRF modelalso showed that the peak time of diurnal cycle of rain-fall was mainly controlled by the deep convective cumulus parametrization, and the planetary boundarylayer scheme could modify the amplitude of the rain-fall diurnal cycle(Koo and Hong, 2010).

Besides the deep convective cumulus parameterization, the large-scale condensation process can alsoproduce precipitation. When total rainfall was dividedinto stratiform and convective rainfall(Yuan, 2013), six GCMs were found to successfully simulate the di-urnal variation of stratiform rainfall with a maximumin the early morning, but most models had problemsin simulating diurnal variations of convective rainfall(Fig. 7 in Yuan, 2013). BCC-CSM1-1 and FGOALS-g2 simulated nocturnal peaks of convective rainfallover all l and areas, but HadGEM2-A, INMCM4, and MRI-AGCM3-2H produced convective rainfall near local noon. The better performance of MRI-CGCM3 isattributed to the well captured ratio of the two kindsof rainfall, but not the diurnal variations of the twocomponents. The diurnal phase of stratiform rainfallis consistent with the diurnal variation of large-scalecirculation, which is also reasonably well simulated byall the models. The results in Yuan(2013)further confirmed that major biases in rainfall diurnal cycles oversubtropical China are associated with the convectiveparameterization. 5. Conclusions and discussion

By summarizing the recent progress in studies ofthe rainfall diurnal variation, this paper presents acomprehensive review on the characteristics and regional differences of rainfall diurnal variations overcontiguous China, and points out the limitations ofcurrent numerical models in simulating rainfall diurnal variations over China. The main conclusions arelisted as follows.

1)The rainfall diurnal variation over contiguousChina shows distinct regional characteristics. Over thesouthern inl and China and northeastern China, summer precipitation peaks in the late afternoon, whileit peaks around midnight over most of the TP and its east periphery. The diurnal phase changes east-ward along the YR valley, with a midnight maximumin the upper valley, an early morning peak in the middle valley, and a late afternoon maximum in the lowervalley. Summer precipitation over the region betweenthe Yangtze and Yellow rivers and most regions of theTP has two diurnal peaks: one in the early morning and the other in the late afternoon.

2)The rainfall diurnal cycle shows a robust seasonal variation, with a dominant rainfall peak in theafternoon(early morning)in warm(cold)seasons. Corresponding to the sub-seasonal movement of monsoon rain belt, the rainfall diurnal variation in warmseason presents distinct sub-seasonal evolution, withthe peak occurring more frequently in the early morning(afternoon)in monsoon active(break)periods. 3)The cloud structure and diurnal phase of long- and short-duration rainfall events show evident differences. The long-duration rainfall is usually characterized by stratiform precipitation. The peak hour of rainrate at surface and highest profile prefer to occur in thelate afternoon, and show an eastward delayed phasetransition along the YR valley. The short-durationrainfall is usually dominated by convective precipitation, with the maximum surface rain rate and the highest profile in the afternoon to early evening.

4)The rainfall diurnal variation is greatly influenced by mountain-valley and l and -sea breezes on various scales. The rainfall tends to reach the diurnalmaximum in the late afternoon(early morning)inthe mountain(valley)regions. The different diurnalregimes are highly correlated with the inhomogeneousunderlying surface, such as the plateau, plain, and ocean, with physical mechanisms consistent with thelarge-scale "mountain-valley" and "l and -sea" breezes and convective instability, resulting in the large-scaleregional differences in the rainfall variation over EastAsia.

5)The diurnal cycle of the lower tropospheric circulation greatly influences the regional differences ofrainfall diurnal variations. The clockwise rotation oflow-level wind is thought to contribute to the eastwarddelay of diurnal phase of long-duration rainfall eventsalong the YR valley, the different diurnal phases between the south and north of the Huai River, and thesecondary peak of summer rainfall over South China.

6)Evaluation of numerical models indicates thatthe capability of current models in simulating therainfall diurnal variation over China is still limited. Increasing the model resolution does not notably improve the model performance. The inability to properly reproduce the regional characteristics in the diurnal variation is primarily due to the problems in themodel physics.

The diurnal variation of precipitation is a widelyconcerned atmospheric phenomenon. The climatic diurnal cycle of rainfall in a specific region implies theinteraction of local integrated forcing with the atmosphere circulation. Studying the rainfall diurnal cycleis an important way to underst and rainfall processes and improve the capability of numerical models inprecipitation forecasting and simulation. However, limited by the observation technology and data, thediurnal variation of rainfall still needs further study.

With the development of modern observationtechnique and continuously enriched high spatiotemporal observational data, the future research on thediurnal variation of rainfall should focus on the following two aspects. Firstly, investigation on the rainfalldiurnal variation should be combined with that on theevolution of cloud systems. To investigate the micro and macro characteristics of clouds related to rain-fall events as well as the three-dimensional structureof rainfall systems will promote the underst and ing ofrainfall evolution features. Li et al. (2003)analyzedthe diurnal changes of stratiform clouds in SouthChina and illustrated the different diurnal variationsof stratiform clouds over South China and the eastof TP. The cloud fraction decreases in the day and increases in the night over coastal regions in SouthChina, while the mid-level stratiform clouds have nodistinct diurnal changes over eastern China. Based onhourly infrared satellite brightness temperature, Chenet al. (2012a)showed that the summer mean frequency of clouds presents a notable diurnal variationover the southern contiguous China. Deep convectionexhibits a large diurnal amplitude in frequency and occurs most frequently in the late afternoon in mostregions. The frequency of mid-level clouds shows arelatively weaker diurnal amplitude and presents adominant nocturnal maximum during the day. Thediurnal variation of different clouds corresponds wellto the different diurnal variations of summer rain-fall between southwestern and southeastern China asrevealed by both station observations and TRMMproducts.

Secondly, future studies should examine the rain-fall diurnal variation in the context of rainfall evolution processes. The current studies mainly reveal and interpret only the diurnal amplitude and phase ofrainfall amount(although a few studies involve rain-fall frequency and intensity), and the association withrainfall processes is still obscure. Yu et al. (2013)revealed that the rainfall process is asymmetric, whichmeans rainfall events usually reach the maximum ina short period and then experience a relatively longerretreat to the end of the event. The results impliedthat there could be a corresponding lead or lag relationship among the diurnal variation of the rainfallamount, frequency, and intensity, rather than the consistent changes as proposed by previous studies. Theresults also suggested the importance of rainfall starting and ending time. For instance, Chen et al. (2010)illustrated the most frequent starting and peakinghours of the warm season long-duration rainfall eventsalong the YR valley. In most regions, the rainfallprefers to initiate in the afternoon and late afternoon and reaches the diurnal maximum quite soon. Alongthe YR valley, the rainfall tends to start in nocturnalhours and peaks after 3-6 h.

Furthermore, on the basis of the full underst and ing of rainfall evolution on diurnal timescale and itsphysical mechanisms, we should also target the workthat closely relates the research on rainfall diurnalvariations with refined weather forecast services. Forexample, detailed assessment of the representationof rainfall diurnal variations in weather predictionmodels, and then the effective error corrections onthe diurnal variation, will warrant more reliable highresolution forecast products. Meanwhile, refined prediction products can also be produced by combiningthe daily rainfall model forecast results with the localrainfall diurnal variation features.