Warm-Season Diurnal Variations of Total, Stratiform, Convective, and Extreme Hourly Precipitation over Central and Eastern China

Precipitation and convective weather both have distinct diurnal cycles. Studies on diurnal precipitation variation can be traced back to the middle 19th century (Wallace, 1975). Numerous previous studies have investigated diurnal variations in precipitation and convective weather in various regions and seasons. It is well recognized that diurnal variations of precipitation and convective weather are different between land and ocean, between mountains/plateaus and valleys/plains and in different seasons. Diurnal variations in precipitation and convective weather are closely associated with factors like solar radiation, land-sea breeze, mountain-valley winds, near-surface friction, nocturnal low-level jets, clouds, and longwave radiation (e.g., Wallace, 1975; Dai, 2001; Mori et al., 2004; Zheng et al., 2008; Chen et al., 2010, Chen et al., 2013, Chen et al., 2014, Chen et al., 2016, Chen et al., 2017, Chen et al., 2018a; Fujita et al., 2010; Huang et al., 2010; Yu et al., 2010; Bao et al., 2011; Yin et al., 2011; Zheng and Chen, 2013; Luo et al., 2016; Jiang et al., 2017; Zhu et al., 2017; Du and Rotunno, 2018; Zhang et al., 2018).

Compiled and quality-controlled hourly precipitation data in China have been available since 2007 (Yu et al., 2007b), and many studies have already investigated the diurnal variations in precipitation in China. (Yu et al., 2014) reviewed the literature, revealing the major conclusions of previous studies to be as follows. The peak time of precipitation in southeastern and northeastern China is concentrated in the late afternoon; the peak time occurs largely in the middle of the night in southwestern China; precipitation occurs more often in the early morning in the middle and upper reaches of the Yangtze River; in central and eastern China, diurnal precipitation variation is characterized by a bimodal feature, with the peak time occurring in the early morning and in the late afternoon, respectively; precipitation in the Tibetan Plateau also shows a bimodal feature, with the peak time in the late afternoon and in the middle of the night, respectively (Yu et al., 2007a, Yu et al., 2007b; Li et al., 2008; Liu and Fu, 2010; Yuan et al., 2010; Xu and Zipser, 2011; Chen et al., 2013). Large differences are also found in the cloud structure and the diurnal peak time between persistent and short-duration precipitation (Yu et al., 2010). Most of the persistent precipitation is stratiform precipitation, and the peak time of the precipitation and precipitation profile largely occurs during the period from midnight to early morning Yu et al., 2007a, 2010). Short-duration precipitation is mostly convective, which often peaks in the late afternoon. (Zhuo et al., 2014) revealed diurnal variations in amount, intensity and frequency of precipitation at various levels in Shandong, indicating the variations are quite different over regions with different topographies. (Yu and Li, 2016) further classified mainland China into seven types of regions based on the peak time of diurnal total-precipitation variations, i.e., regions with their precipitation peak time in the late afternoon [mountainous region from Northeast to North China and the inland region of Southeast China, i.e., regions AN_N and AN_S in the study of (Yu and Li, 2016)], regions with their precipitation peak time at nighttime [western North China Plain near the mountains and from the western Sichuan Basin to the eastern Yunnan-Guizhou Plateau, i.e., regions MN_N and MN_S in the study of (Yu and Li, 2016)], regions with their precipitation peak time in the early morning [eastern North China Plain, the Shandong Peninsula, the Liaodong Peninsula, and from the Qin-Ba mountainous region to southwestern Central China, i.e., regions EM_N and EM_S in the study of (Yu and Li, 2016)], and the Tibetan Plateau, where the precipitation peak time occurs in the evening.

Some other studies (Zhang and Zhai, 2011; Chen et al., 2013) revealed the characteristic diurnal variations in extreme hourly precipitation or short-duration heavy precipitation. Their results indicated that these variations also demonstrate various features of unimodal, bimodal, multimodal, and persistently active periods over different regions. In addition, the peak time periods of diurnal variation in short-duration heavy precipitation are also different over different underlying surfaces. (Zhou et al., 2008) compared the diurnal variations in precipitation amount, intensity and frequency from rain-gauge observations and from retrievals of satellite observations, and found that the results from the two types of data are consistent with each other. Nevertheless, biases are also found in precipitation retrievals from satellite observations. (Chen et al., 2013) also found that the diurnal variation characteristics of short-duration heavy precipitation from rain-gauge observations are quite consistent with those of mesoscale convective systems (MCSs) retrieved from satellite observations, except at 0200-0800 LST (Local Standard Time, UTC+8).

Despite the great achievements in studying diurnal precipitation variation in China, there still exist some weaknesses. Most previous studies used observations of fewer than 700 stations (e.g., Yu et al., 2007a, Yu et al., 2007b, Yu et al., 2010; Zhang and Zhai, 2011; Chen et al., 2013) in China, except those of (Zhuo et al., 2014) and (Yu and Li, 2016), who implemented high-density observations to analyze the diurnal variations in precipitation in Shandong and over the whole of China, respectively. Observations collected at weather stations that are sparsely distributed cannot fully capture the variational features of meso- and microscale convective precipitation, which may be extreme (Zheng et al., 2016). For this reason, precipitation retrievals of satellite data have been used in some studies. However, as indicated by (Zhou et al., 2008) and (Chen et al., 2013), there are certain differences between retrievals of satellite data and rain-gauge observations.

Although it has been recognized that large differences exist in the diurnal variations between precipitation at different levels and with different durations, and between convective and stratiform precipitation (e.g., Wallace, 1975; Dai, 2001; Yu et al., 2007a, Yu et al., 2010; Liu and Fu, 2010; Zhuo et al., 2014), comprehensive analyses of diurnal variations in hourly precipitation at various levels for total, stratiform, convective, and extreme precipitation in China have not been conducted yet. Specifically, the relationship between diurnal variations in convective and stratiform precipitation with different intensities still remains unclear. Due to the limitations in Tropical Rainfall Measuring Mission (TRMM) data, comparative studies on the diurnal variations in convective and stratiform precipitation only give results over the region south of 35^°N in China (e.g., Liu and Fu, 2010; Yu et al., 2010). Note that there is no definite threshold of hourly precipitation to completely distinguish stratiform precipitation from its convective counterpart (e.g., Wallace, 1975; Steiner et al., 1995). However, (Steiner et al., 1995) pointed out that stratiform precipitation intensity is seldom greater than 10 mm h^-1. Therefore, one of our aims is to obtain the diurnal variations with different precipitation intensities to show whether their characteristics are consistent with the revealed features of stratiform precipitation or convective precipitation (e.g., Yu et al., 2010; Zhang and Zhai, 2011; Chen et al., 2013).

Although (Yu and Li, 2016) classified mainland China into seven types of regions based on diurnal total-precipitation variations, no study using diurnal variations of different types of precipitation to classify the regions has been performed. In addition, extreme hourly precipitation no less than 50 mm occurs so rarely (Zhang and Zhai, 2011; Luo et al., 2016; Zheng et al., 2016) that little is known about its diurnal variation because of the small number of samples of such precipitation events and limited data availability.

To address the issues mentioned above, the present study analyzes the diurnal variations of hourly precipitation and their similarities and differences at various levels (including total, stratiform, convective, and extreme hourly precipitation) over various regions of central and eastern China. The results of the present study provide a climatological basis for understanding the mechanisms of hourly precipitation with different intensities, evaluating cumulus parameterizations and other microphysical processes in weather and climate models (Yu et al., 2007a,b), and improving the capacity for forecasting such precipitation events.

The hourly precipitation dataset used in the present study was collected at national weather stations in China, and is the same as the one used in the studies of (Luo et al., 2016) and (Zheng et al., 2016). The dataset is provided by the National Meteorological Information Center of the China Meteorological Administration. It covers the period 1951-2012 and has gone through strict quality control (Luo et al., 2016; Zhang et al., 2016; Zheng et al., 2016). The number of valid observations in the dataset varies with time, and the number of stations is fewer than 1000 in the 1950s (Zhang et al., 2016; Zheng et al., 2016). In this paper, only observations at stations with less than 80% of the total observations that should be collected are excluded to take advantage of the high-resolution observations and better represent the hourly precipitation climatology. Only those observations that are available for no less than 30 years during the period of May-September of 1960-2012 are selected for analyses. Based on the above criteria, observations at 2061 stations can be used for the present study. The spatial distribution of these stations is uneven (Fig. 1), and the average distance between the stations is about 50 km. Figure 1 shows that most of these stations are located in central and eastern China. Therefore, the present study only focuses on the diurnal variations of hourly precipitation in central and eastern China.

Figure 1.  Topography and weather stations used for the present study. Shaded areas indicate topography. Purple and black dots indicate stations from which the observations are used, and black dots labeled in smaller black letters indicate different regions where diurnal variations are presented in the study, respectively. Thick blue lines denote the boundaries of various larger regions labeled with large red letters, and cyan lines represent the locations of cross sections. The inset in the lower left represents the South China Sea Islands.

Seven hourly precipitation thresholds, i.e., 0.1, 0.5, 1, 5, 10, 20, and 50 mm, are specified to investigate the diurnal variations in precipitation amount, frequency, and intensity for hourly precipitation no less than these thresholds, respectively. Hourly precipitation of 0.1 mm represents the threshold that determines the status of precipitation or non-precipitation (e.g., Wallace, 1975; Yu et al., 2007b; Yu and Li, 2016). As mentioned above, no definite threshold of precipitation intensity can be used to fully distinguish the stratiform component from the convective component (e.g., Wallace, 1975; Steiner et al., 1995). There is also no absolute threshold to determine precipitation extremes. So, (Chen et al., 2018b) used the precipitation fractional coverage to classify the mei-yu precipitation into two weather regimes: large coverage and small coverage. Nevertheless, hourly precipitation of less than 2.5 mm usually represents stratiform precipitation in the USA (Wallace, 1975), and (Fu and Liu, 2003) and (Liu and Fu, 2010) found that mean hourly stratiform precipitation derived from TRMM observations is 2-3 mm in East Asia, which is nearly equal to the threshold of stratiform precipitation given by (Wallace, 1975). The study of (Steiner et al., 1995) revealed that over the neighboring area of Darwin in Australia, both mean and peak-frequency stratiform precipitation derived from radar observations are only about 1 mm h^-1, and the precipitation seldom exceeds 10 mm h^-1. As proposed by (Wallace, 1975), "trace" precipitation (<2.5 mm h^-1) events can only be interpreted ambiguously, because these events may be associated with different types of precipitation, such as "prolonged periods of light drizzle, snow flurries, sprinkles from weak showers, or from more vigorous storms that don't pass directly over the station, etc." (Wallace, 1975). Nonetheless, (Wallace, 1975) still proposed that the diurnal variation in the frequency of "trace" precipitation events "is the most sensitive to variations in light precipitation from stratiform clouds" (Wallace, 1975). Therefore, besides 0.1 mm h^-1, 0.5 mm h^-1 and 1 mm h^-1 are also selected as thresholds, mainly to identify heavier stratiform precipitation and further show their similarities and differences in diurnal variation. It is important to note that the diurnal variations of precipitation at the levels ≥ 2 mm h^-1, ≥ 2.5 mm h^-1, and ≥ 3 mm h^-1 (figures not shown) are very similar to that at the level ≥ 1 mm h^-1, which indicates that the variation in stratiform precipitation is insensitive to the threshold values between 1 mm h^-1 and 3 mm h^-1. Hourly precipitation of 5 mm and 10 mm could be either stratiform precipitation or convective precipitation, or a mix of both (Steiner et al., 1995; Fu and Liu, 2003; Chen et al., 2013). Note that, in the arid or semi-arid regions of inland China, 5-mm hourly precipitation might be produced by convection. Hourly precipitation up to 20 mm and above is defined as short-duration heavy precipitation, which can only be convective (Davis, 2001; Fu and Liu, 2003; Zhang and Zhai, 2011; Chen et al., 2013). Such precipitation is prone to result in flooding (Brooks and Stensrud, 2000; Davis, 2001). Extremely heavy precipitation of no less than 50 mm h^-1 must be produced only by convection (Davis, 2001; Zhang and Zhai, 2011; Chen et al., 2013; Li et al., 2013a, b; Luo et al., 2016; Zheng et al., 2016).

The approach proposed by (Yu and Li, 2016) is used in the present study to calculate precipitation amount, frequency and intensity, i.e., precipitation amount during a specific period for a specific threshold is the accumulated hourly precipitation that is no less than the threshold; precipitation occurrence number is the accumulated number of hours when hourly precipitation is no less than a specified threshold; precipitation occurrence frequency is the ratio of precipitation occurrence number to total observation number. At each hour of 0000-2300 LST, precipitation intensity is defined as the accumulative precipitation divided by the accumulative precipitation occurrence number at each individual specified threshold.

In order to compare the diurnal variations in hourly precipitation at different levels, the precipitation amount, frequency, and intensity are normalized following the approach of (Yu and Li, 2016), i.e., hourly values are divided by their corresponding 24-h average. By normalization, diurnal variations in precipitation amount, frequency, and intensity at different levels over various regions can be displayed in the same coordinate system and easily compared with each other.

As is well known, distinct differences exist in diurnal precipitation variations over different regions. Thereby, based on the classified precipitation regions of (Yu and Li, 2016), the present study divides central and eastern China into nine regions (Fig. 1). Region A mainly includes Northeast China; Region B covers most of Inner Mongolia, northern Shannxi, most of Shanxi, and northern Hebei, which is the same as the region AN_N in the study of (Yu and Li, 2016); Region C is the same as region MN_N in the study of (Yu and Li, 2016) and contains southern Beijing, western Hebei and northern Henan; Region D is consistent with the region EM_N in the study of (Yu and Li, 2016) and covers part of Henan, the northern part of both Jiangsu and Anhui, Shandong, and southern Liaoning; Region E is region MN_S in the study of (Yu and Li, 2016), which mainly includes western Sichuan and western Guizhou; Region F is the same as region EM_S in the study of (Yu and Li, 2016) and includes southern Shannxi, eastern Sichuan, Chongqing, eastern Guizhou, and most of Hunan; Region G includes southwestern Henan, most of Hubei, the southern part of both Jiangsu and Anhui, and northwestern Jiangxi; Region H is mainly located in Yunnan and western Guangxi; Region I is region AN_S in the study of (Yu and Li, 2016), which covers most of Jangxi and Zhejiang, the whole of Fujian, eastern Guangxi, Guangdong, and Hainan Island.

The diurnal precipitation peak time at certain weather stations in each type of region classified by (Yu and Li, 2016) might be in opposite phase. Thereby, several weather stations in different regions are selected in the present study to illustrate multiple types of diurnal precipitation variation. These regions include the Northeast China Plain (A-P), the Loess Plateau (B-G), northern North China Plain (C-P), the Liaodong Peninsula (D-L), the Shandong Peninsula (D-S), the Sichuan Basin (E-S), the middle reaches of the Yangtze River (G-C), Shanghai and surrounding area (G-S), the Hengduan Mountain area in Yunnan (H-M), the plateau area near Kunming (H-G), and the coastal regions of Hainan Island (I-H) and Guangdong (I-S), respectively. These selected stations are denoted by black dots in Fig. 1. Furthermore, longitude-time and latitude-time cross sections (Fig. 1) of precipitation are plotted to display the diurnal variations in precipitation at two specific levels to present the differences in propagation and diurnal variation between total and convective precipitation. In order to increase the sample size, precipitation observations within 1^° to the north and south (east and west) of each selected parallel (meridian) are used for producing the cross section plots.

In order to understand the overall characteristics of diurnal precipitation variation over central and eastern China, averaged diurnal variations in hourly precipitation amount, frequency, and intensity at various levels for the warm season are presented in Fig. 2, which shows that large differences exist in the diurnal variation of precipitation at different levels. To further understand the differences in the average precipitation amounts, occurrence frequencies, and intensities at different levels over the whole of central and eastern China and the various sub-regions, those averaged values are listed on the right-hand side of Fig. 2, and in other figures.

For hourly precipitation at the level ≥ 0.1 mm representing total precipitation, the diurnal variations in amount and frequency (dashed lines in Fig. 2) are completely consistent with the results of (Yu and Li, 2016). The variations in both amount and frequency show a bimodal feature, with the largest precipitation amount occurring at 1600-1700 LST and 0600-0700 LST, respectively; and the peak precipitation amount that occurs in the late afternoon is slightly larger. However, the peak precipitation frequency appears at 0700 LST, and the secondary peak appears at 1700 LST. The diurnal variation in precipitation intensity demonstrates a unimodal pattern, with the peak value occurring at 1600-1700 LST, corresponding to the peak time periods of amount and frequency of hourly precipitation ≥ 20 mm that represents convective precipitation. These overall features reflect the fact that precipitation amount and intensity in the late afternoon are both large over central and eastern China, which is caused by convective precipitation. In contrast, stratiform precipitation dominates after midnight, and it is generally weak, although it occurs more frequently at nighttime than in the late afternoon.

The diurnal variations between precipitation amount and frequency shown in Fig. 2b and 2c are quite similar for the same level precipitation. The diurnal variations in all the precipitation at levels other than that ≥ 0.1 mm h^-1 show a bimodal pattern, with the primary peak occurring in the late afternoon and the secondary peak occurring after midnight, and the normalized amplitude of the variation increases with the increasing threshold as the thresholds are greater than 1 mm h^-1. This is because convective precipitation over land mainly occurs in the late afternoon (e.g., Wallace, 1975; Zheng et al., 2008; Yu et al., 2010; Chen et al., 2013; Yu and Li, 2016). Figure 2 also shows that the amplitudes of the diurnal variations in precipitation frequencies at the levels ≥ 0.5 mm h^-1 and ≥ 1 mm h^-1 are significantly smaller than those at other levels, i.e., the occurrence frequencies of precipitation at each level of the two for each hour are by and large closer than those at other levels. The diurnal variation in precipitation at the level ≥ 5 mm h^-1 is already close to that at the level ≥ 20 mm h^-1, suggesting that precipitation above the level ≥ 5 mm largely consists of convective precipitation, but it also contains some stratiform precipitation. The diurnal variation in precipitation at the level ≥ 10 mm h^-1 demonstrates the variational feature of convective precipitation, which is quite similar to that of convective precipitation at the level ≥ 20 mm h^-1. These results indicate that the precipitation at the level ≥ 10 mm is mainly convective, and the effect of stratiform precipitation on the diurnal variation can be ignored. This is consistent with the result of (Chen et al., 2013), who showed that the climatological pattern of precipitation at the level ≥ 10 mm h^-1 is quite similar to that at ≥ 20 mm h^-1 in China. The diurnal variation in precipitation at the level ≥ 20 mm h^-1 is also similar to that in (Chen et al., 2013), although the pattern is slightly different. This is because the hourly precipitation data used by (Chen et al., 2013) covered only a short period and observations were available only at a limited number of stations. For this reason, the diurnal variation shown in their work is less continuous and smooth. Because precipitation at the level ≥ 50 mm h^-1 is extremely heavy precipitation, its diurnal variation features a large amplitude with the primary peak occurring at 1700 LST, and the secondary peak occurring at 0200 LST; and the variation between 2000 LST and 0800 LST is mild, indicating that the precipitation amount and frequency change little during this period.

Figure 2.  Averaged normalized diurnal variations in (a) intensity, (b) frequency, and (c) amount of precipitation at various levels over the whole of central and eastern China. The x-axis indicates time, in LST. Color-shaded areas indicate diurnal precipitation variations at different levels. The white solid lines are the isolines with an average value of 1. The y-axis on the left indicates hourly precipitation at different unevenly-spaced levels (units: mm). The numbers on the right-hand side indicate the average values (the units for precipitation amount and intensity are mm, and the units for precipitation occurrence frequency are %). Specifically, diurnal variations in precipitation at the levels ≥0.1 mm and ≥ 20 mm are presented by dashed and solid lines, respectively; the y-axis on the left indicates the normalized diurnal variation, and the x-axis indicates time, in LST.

Figure 2a shows that the diurnal variation in precipitation intensity is significantly different to the variations in precipitation amount and frequency. Precipitation intensity displays insignificant diurnal variation, except that at the level ≥ 0.1 mm h^-1, which is attributed to the Γ distribution of hourly precipitation frequency (Tian et al., 2014). The diurnal variation in precipitation intensity at the level ≥ 0.1 mm h^-1 fully agrees with that of (Yu and Li, 2016), showing a unimodal feature with the peak occurring in the late afternoon. As stated earlier, this is determined by the peak frequency of convective precipitation, which contributes to the late-afternoon peak of precipitation amount. For precipitation at other levels, although they all show an insignificant unimodal feature, some differences exist in peak time and amplitude, i.e., with an increase in the threshold, the peak time delays significantly, the peak value and amplitude decrease greatly, the unimodal pattern becomes less obvious, and the diurnal variation weakens, indicating diurnal variations between precipitation amount and frequency become much more similar. For precipitation at the level ≥ 50 mm h^-1, the precipitation intensity during 2000-1200 LST is slightly larger than that during 1200-2000 LST, suggesting that the precipitation intensity is stronger at nighttime than in the late afternoon from the perspective of climatological statistics, although extremely heavy precipitation occurs less frequently at nighttime than in the late afternoon (shown in Fig. 2b). This is possibly because the diurnal peak of extremely heavy precipitation at the level ≥ 50 mm h^-1 mostly occurs at nighttime in Southwest China (regions E, F, and H) and South China (region I).

On the basis of the occurrence time of the primary peak frequency of precipitation at various levels, the nine regions shown in Fig. 1 are further classified into three categories, i.e., regions where the primary peak appears in the late afternoon, regions with a nocturnal primary peak, and regions with a shifting primary peak time.

4.1.1. Diurnal variations with a late-afternoon primary peak

Due to surface heating by solar radiation, convection over land is at its most vigorous in the late afternoon (e.g., Wallace, 1975; Dai, 2001; Zheng et al., 2008; Chen et al., 2013; Zheng and Chen, 2013; Yu and Li, 2016). Among the nine regions shown in Fig. 1, two regions (B and I) present a diurnal precipitation variation feature with a frequency peak appearing in the late afternoon for all precipitation levels (Fig. 3). These regions are regions AN_N and AN_S given by (Yu and Li, 2016), where the diurnal precipitation variations differ widely from the overall pattern shown in Fig. 2.

The diurnal variation features of precipitation amount and frequency over region B (Fig. 3a) are as follows. For precipitation at the levels ≥ 0.1 mm h^-1, ≥ 0.5 mm h^-1, and ≥ 1 mm h^-1, their diurnal variations are similar and feature a bimodal pattern, with the primary frequency peak appearing in the late afternoon and the secondary frequency peak in the early morning. For precipitation at the level ≥ 5 mm h^-1, the diurnal precipitation variation shows a significant unimodal pattern, which is typical for convective precipitation. This result indicates that the primary peak reflects the feature of convective precipitation, while the secondary peak presents the feature of stratiform precipitation. As convective precipitation occurs more frequently and more intensely in the late afternoon, total precipitation amounts are much higher than those in the early morning.

Figure 3.  As in Fig. 2 but for regions (a) B and (b) I.

Over region I (Fig. 3b), the troughs of the diurnal variations in precipitation amount and frequency both occur between 2300 and 0100 LST, which are different to those over region B. In other words, precipitation amount and frequency in the morning are larger than those in the middle of the night, which is associated with the geographic location of region I being in southeastern China, where the climatological and geographical features are unique. Note that the diurnal variation over region I-S along the coastal region of South China, which will be presented and further discussed later, is distinctly different from that over region I. The diurnal variation in precipitation at the level ≥ 50 mm h^-1 shows a bimodal pattern, with the primary peak occurring in the late afternoon and the secondary peak in the early morning. An extremely heavy precipitation event with hourly rainfall of 184.4 mm that occurred between 0500 and 0600 LST on 7 May 2017 in Guangzhou is a typical nocturnal case (Tian et al., 2018).

The diurnal variations in precipitation intensity over regions B and I change from unimodal to an insignificant pattern with an increasing threshold of precipitation level. This result indicates that nocturnal precipitation is usually weaker than that in the late afternoon. However, for extreme precipitation at the level ≥ 50 mm h^-1 over region I, the intensity is slightly heavier after midnight than in the late afternoon.

4.1.2. Diurnal variations with a nocturnal primary peak

Many previous studies have revealed that precipitation and convection in Southwest China, especially the Sichuan Basin, mainly occur at nighttime (e.g., Yu et al., 2007a, b; Li et al., 2008; Zheng et al., 2008; Yu et al., 2010; Yuan et al., 2010; Xu and Zipser, 2011; Chen et al., 2013; Yu and Li, 2016). This is related to the terrain in this region. However, except for the studies of (Chen et al., 2013), (Yu et al., 2010), and (Yu and Li, 2016), little attention has been paid to nocturnal precipitation over the Yunnan-Guizhou Plateau. In the present study, Southwest China is divided into regions E, F, and H. The primary peaks of diurnal variations in precipitation amount and frequency all appear at nighttime (Fig. 4), and all the peaks of convective precipitation at nighttime are ahead of their stratiform precipitation counterparts. This result basically agrees with the features of diurnal variations of stratiform and convective precipitation over Southwest China presented in the study of (Yu et al., 2010), which was based on TRMM data. However, (Yu et al., 2010) did not reveal the fact that the heavier the convective precipitation, the earlier the peak of the convective precipitation occurs (Fig. 4).

Figure 4.  As in Fig. 2 but for regions (a) E, (b) F and (c) H.

Over region E, the diurnal variations in both precipitation amount and frequency show a unimodal feature, with the peak appearing between midnight and early morning. In comparison, most of the diurnal variations over regions F and H present a bimodal feature, especially for all the precipitation levels no less than 5 mm h^-1, most of which must be produced by convection. This result also reflects the difference in diurnal variation between convective and stratiform precipitation, suggesting that in these regions, stratiform precipitation mainly occur between midnight and morning, while convective precipitation largely occurs after midnight or in the late afternoon. Note that the phases of the secondary peaks of precipitation amount and frequency over regions F and H also change with precipitation level, and the peaks of stratiform precipitation in the late afternoon occur before those of convective precipitation. This is possibly because stratiform precipitation usually lasts for a long time in these two regions, while convective precipitation occurring in the late afternoon only lasts for a short period (Zheng et al., 2008, Zheng et al., 2016; Yu et al., 2010).

The diurnal variations in precipitation intensity over regions E and F show that the precipitation at most levels is heavier at nighttime than in the daytime, except for precipitation at the level ≥ 50 mm h^-1, whose intensity changes insignificantly with time. In contrast, over region H, the diurnal peaks of precipitation intensity at the levels no greater than 5 mm h^-1 occur between late afternoon and early morning, while those at the levels ≥ 10 mm h^-1, ≥ 20 mm h^-1, and ≥ 50 mm h^-1 appear between evening and morning. This feature is also determined by the active time period of convective precipitation.

4.1.3. Diurnal variations with a time-shifting primary peak

Over regions A, C, D, and G (Fig. 5), the diurnal variations in stratiform precipitation frequency are quite different from the variations in convective precipitation frequency; furthermore, the patterns over different regions are also somewhat different from each other. Over these regions, the primary peak time period of precipitation frequency changes as the precipitation level threshold increases.

Figure 5.  As in Fig. 2 but for regions (a) A, (b) C, (c) D and (d) G.

The diurnal variation features of precipitation amount and frequency over region A are similar to those over region B, as analyzed above. However, the diurnal variations in total and stratiform precipitation frequency, which feature a bimodal pattern with two almost-the-same amplitude peaks, are different from their counterparts over region B. Regions C, D, and G (Fig. 1) correspond to the regions where the peak of total precipitation frequency mainly appears at night or in the early morning in the study of (Yu and Li, 2016), although there are a few stations with opposite-phase peak time. Over regions C and D, the diurnal variations in precipitation amount and frequency at various levels all display a bimodal feature, and stratiform precipitation occurs during the period from late afternoon to early morning over region C and mainly in the early morning over region D, but convective precipitation primarily occurs in the late afternoon. Note that over region C, the peaks of convective precipitation at nighttime are earlier than those of its stratiform counterpart, which is similar to the case over regions E, F, and H. Over region G, except for precipitation at the levels ≥ 20 mm h^-1 and ≥ 50 mm h^-1 with no nocturnal peak, the diurnal precipitation variations at all other levels show one peak in the late afternoon and the other peak in the early morning. Similar to the pattern in region B, all the diurnal variations in precipitation amount over regions A, C, and G show a primary peak in the late afternoon, which is also attributable to the fact that convective precipitation over these regions occurs more frequently and more intensely in the late afternoon. However, over region D, the diurnal variations in total and stratiform precipitation amount are quite different from their counterparts over regions A, C, and G, and this is because the convective precipitation at the levels ≥ 5 mm h^-1, ≥ 10 mm h^-1, and ≥ 20 mm h^-1 occurs more frequently in the early morning (shown in the middle panel of Fig. 5c).

Over these four regions, Fig. 5 shows that the peak time periods of total precipitation intensity are quite consistent with the peak time periods of precipitation frequency at the level ≥ 10 mm h^-1. Similar to the overall diurnal variations presented in section 3, the diurnal variations in total precipitation intensity are determined by convective precipitation.

As shown in the study of (Yu and Li, 2016), the peaks of precipitation amount and frequency at some stations might be in opposite phase to that over the whole region with a pattern of diurnal variation. (Zheng and Chen, 2013) pointed out that the coastal region of South China is a transitional zone of diurnal variation in strong convective activity, where convective activities have longer active periods than in the surrounding ocean and land. Thereby, several stations within a small area of the regions classified in section 4.1 are selected for further analyses of their diurnal variation features to show that different physical processes can lead to different diurnal variation features. Multiple stations instead of a single station are selected for the present study, mainly because the samples of precipitation at the level ≥ 50 mm h^-1 at one single station are quite limited, which makes it hard to illustrate its diurnal variation feature.

Diurnal variations at typical stations in regions A-P, B-G, C-P, E-B, G-S, H-G, H-M, and I-H (figures not shown) are quite similar to the overall diurnal variations over the larger region where these smaller regions are located when considering the peak pattern and peak time period, although some differences exist in the amplitude and duration of the peak. In contrast, in regions D-L, D-S, G-C, and I-S (Fig. 6), the diurnal variations at the typical stations are considerably different to the overall features over their corresponding larger regions given in section 4.1.

Figure 6.  As in Fig. 2 but for the weather stations over regions (a) D-S, (b) D-L, (c) G-C and (d) I-S. The stations are denoted by black dots in Fig. 1, and the areas shaded in black indicate there is no observation.

Regions D-L, D-S, I-S, and I-H are all located in the coastal region of China, and region G-C is located in the middle reaches of the Yangtze River, where total and stratiform precipitation occurs more frequently in the early morning. The diurnal variations over region G-C (Fig. 6c) show that the stratiform precipitation occurs from midnight to evening, with a peak in the early morning, and the convective precipitation at the levels ≥ 10 mm h^-1, ≥ 20 mm h^-1 and ≥ 50 mm h^-1 exhibits two peaks with almost the same amplitudes occurring in the early morning and in the late afternoon, respectively. This result shows that some mechanisms must exist, like a low-level jet and interactions with existing MCSs (Luo et al., 2014), that support the development of the early-morning MCSs. In comparison, the diurnal variations over regions D-L, D-S, and I-S (Fig. 6) are to a certain degree similar to each other, with the troughs of precipitation amount and frequency both appearing between evening and midnight. Nevertheless, the peaks appear between midnight and early morning over regions D-L and D-S, and between morning and late afternoon over region I-S. However, the variations over region I-H (figure not shown) demonstrate a typical feature of thermal convection, despite the fact that the stratiform precipitation also has a secondary peak of frequency in the early morning. The diurnal precipitation variations over regions I-S and I-H are similar to the diurnal variations in convective activity revealed by (Zheng and Chen, 2013), which are based on blackbody temperature data from geostationary meteorological satellites. This result indicates that the diurnal precipitation variations in these two regions are determined by convective activities.

This section further presents multiple patterns of diurnal variations and propagations of total and convective precipitation, and their similarities, differences, and relationships. The above analyses and results in the literature (Brooks and Stensrud, 2000; Davis, 2001; Chen et al., 2013) all indicate that the diurnal variation in precipitation at the level ≥ 20 mm h^-1 can completely represent the variation in convective precipitation. Therefore, this section only shows several cross sections (their locations are shown in Fig. 1) of diurnal variations in precipitation at the levels ≥ 0.1 mm h^-1 and ≥ 20 mm h^-1, which correspond to total and convective precipitation, respectively.

The averaged latitude-time and longitude-time cross sections of diurnal variations in precipitation intensity, frequency, and amount (Fig. 7), from all available stations, further display the diurnal variation features of the three major patterns as presented in section 4.1, including the pattern with primary peak appearing in the late afternoon, the bimodal pattern with two peaks in the early morning, and in the late afternoon, respectively, and the unimodal pattern with a nocturnal peak. Note that, for the same pattern, some differences still exist in the phase and duration of the peaks.

Figure 7.  Averaged (a) latitude-time and (b) longitude-time cross sections from all available stations. Left-hand panels in (a) and (b) show the diurnal variations for precipitation at the level ≥ 0.1 mm; right-hand panels are for precipitation at the level ≥ 20 mm. From top to bottom in (a) and (b) are diurnal variations in precipitation intensity, frequency, and amount. The white solid lines are the isolines of 1, and areas shaded in black indicate there is no observation.

In Fig. 7, the diurnal variation in amount and frequency of total precipitation mainly features a bimodal pattern, with two peaks usually appearing in the early morning and in the late afternoon, respectively, while that for convective precipitation generally features a pattern with the primary peak appearing in the late afternoon. Over 103^°-107^°E, total and convective precipitation both show a diurnal pattern with peaks of precipitation amount and frequency appearing at nighttime, which reflect the overall feature of diurnal precipitation variations over Southwest China (regions E, F, and H; Fig. 4). In addition, Fig. 7a shows that the primary peak of precipitation near 21^°N occurs in the morning, which reflects the diurnal precipitation variation feature in the coastal region of South China, which is basically consistent with that over region I-S (Fig. 6d) and that near 21^°N in the latitude-time cross section along 110^°E (Fig. 8a).

Figure 8.  As in Fig. 7 but for latitude-time cross sections along (a) 110^°E and (b) 120^°E.

Time-latitude cross sections along 105^°E (figure not shown), 110^°E (Fig. 8a), 115^°E (figure not shown), and 120^°E (Fig. 8b), and time-longitude cross sections along 22^°N (Fig. 9a), 26^°N (figure not shown), 30^°N (Fig. 9b), 35^°N (figure not shown), and 40^°N (figure not shown), respectively, show that the diurnal variations over most parts of each cross section are consistent with those over their corresponding larger region given in the previous sections.

Figure 9.  As in Fig. 7 but for longitude-time cross sections of diurnal precipitation variation along (a) 22^°N and (b) 30^°N.

However, some areas still exist where the diurnal variations are different from those over their corresponding larger regions. Both total and convective precipitation in the area of 34^°-40^°N (located in region B) along 110^°E (Fig. 8a) present a bimodal pattern with a primary peak in the late afternoon, and the diurnal variation pattern of total precipitation is similar to that over region B, whereas the convective precipitation presents a diurnal pattern different from the unimodal pattern over region B. The diurnal variation of convective precipitation, with two peaks of similar amplitude in the area 24^°-30^°N along 110^°E located in region F, is also somewhat different from that over region F. It is worth noting that the diurnal variations over the area 35^°-40^°N along 120^°E, and the area of 106^°-110^°E along 22^°N, all present an obvious peak after midnight, which also differ from those over their corresponding larger regions. These regions are located in the coastal regions of the Bohai Sea or the Beibu Gulf, and the land-sea breeze must play an important role in the variations (Zheng and Chen, 2013; Chen et al., 2016, Chen et al., 2017; Jiang et al., 2017; Zhu et al., 2017; Du and Rotunno, 2018).

Similar to the feature of precipitation intensity displayed in Fig. 2, Figs. 7, 8, and 9 also show that the diurnal variations in total precipitation intensity are significant, whereas the variations in convective precipitation are insignificant. Comparing the variations in the two types of precipitation (Figs. 7, 8, and 9), it can also be found that the occurrence periods of peak amount and frequency of convective precipitation basically correspond to the occurrence time periods of the peak of total precipitation intensity. This result further suggests that convective precipitation determines the total precipitation intensity, since convective precipitation is usually intense.

The time-latitude and time-longitude cross sections also display obvious diurnal precipitation propagation features. The precipitation frequency and amount over the area 21^°-27^°N in Fig. 7a features a northward propagation in the afternoon and a southward propagation after midnight, and an eastward propagation over the area 103^°-111^°E in Fig. 7b. Significant eastward propagation of precipitation from midnight to late afternoon is also found over 106^°-111^°E along 22^° N (Fig. 9a), and along 26^°N (figure not shown). Similarly, precipitation also propagates significantly eastward from midnight to early morning over 103^°-109^°E along 30^° N (Fig. 9b). It is also found that precipitation propagates northward after noon over 22^°-25^°N along 115^°E (figure not shown). The above results agree with those revealed by (Yu et al., 2007b), (Zheng et al., 2008), (Chen et al., 2013), (Zheng and Chen, 2013), and (Jiang et al., 2017). However, convective precipitation (not total precipitation) propagates southward after midnight over 31^°-37^°N along 115^°E (figure not shown), and convective precipitation propagates eastward from midnight to late morning over 111^°-115^°E along 30^°N (Fig. 9b). The above results have not been revealed previously in the literature.

Diurnal variation of precipitation is an important aspect of climatology, and has been paid much attention in China. Nonetheless, (Zhou et al., 2008) and (Yu and Li, 2016) only presented diurnal variations of total precipitation intensity, frequency, and amount; (Zhang and Zhai, 2011) and (Chen et al., 2013) only showed diurnal variations of short-duration heavy precipitation frequency; and (Yu et al., 2010) presented the similarities and differences in diurnal variations between stratiform and convective precipitation only in southern China. Numerous studies (e.g., Yu et al., 2007b; Zhou et al., 2008; Zhang and Zhai, 2011; Chen et al., 2013, Chen et al., 2014, Chen et al., 2016, Chen et al., 2017; Luo et al., 2016; Jiang et al., 2017; Zhu et al., 2017; Du and Rotunno, 2018) have investigated total or short-duration heavy precipitation diurnal variations over several typical subregions of China, but they did not classify these regions.

To the best of our knowledge, the present study is the first to provide the overall characteristics of the diurnal variations in precipitation intensity, frequency, and amount at different levels and show their similarities and differences for different types of precipitation over central and eastern China. Specifically, the variations of precipitation ≥ 50 mm h^-1 have never been revealed in the literature. Both (Wallace, 1975) and (Yu et al., 2010) pointed out that the convective precipitation frequency displays a much larger normalized amplitude of diurnal variation than its total or stratiform counterpart, but the present study is the first to reveal that the amplitudes of normalized diurnal variations of convective precipitation amount and frequency generally increase with an increasing threshold, and convective precipitation determines the variation peak of total precipitation intensity. Although this study cannot fully distinguish convective precipitation from its stratiform counterpart based only on the thresholds of hourly precipitation, the diurnal variations at different levels reveal that the convective precipitation threshold (5 mm h^-1) in northern China is weaker than that (10 mm h^-1) in southern China. We also obtain another fact that the diurnal variations in precipitation frequency at the levels ≥ 0.5 mm h^-1 and ≥ 1 mm h^-1 have the smallest normalized amplitudes.

Although (Yu and Li, 2016) were the first to classify mainland China based on the diurnal variation of total precipitation, the present study has revealed that large differences still exist in the diurnal variations of different types of precipitation in individual regions, and hence the regions have been further classified into three main categories. Furthermore, it is found in this study that the diurnal variations over the coastal regions of the Bohai Sea and South China are one special type, different from the three main categories.

The characteristics of diurnal variations of stratiform and convective precipitation in southern China (Fig. 3b and Fig. 4) are basically identical to the results revealed by (Yu et al., 2010), which demonstrates that the results obtained in this study are reliable. Furthermore, the present study reveals more detailed characteristics of diurnal variations in various regions. For example, precipitation with different intensities over region E all exhibit one peak; the variations over regions F and H have a secondary peak in the late afternoon; in Southwest China and in western North China Plain, the peak of convective precipitation occurs earlier with increasing precipitation intensity at nighttime.

Many studies (e.g., Yu et al., 2007b; Zhang and Zhai, 2011; Chen et al., 2013, Chen et al., 2014, Chen et al., 2016, Chen et al., 2017; Jiang et al., 2017; Du and Rotunno, 2018) have documented the propagation characteristics of total precipitation or short-duration heavy precipitation in several regions of China; however, this study presents the differences in propagation between total and convective precipitation in some regions, not previously revealed in the literature.

Convective precipitation over land in the late afternoon is often triggered by solar radiative heating at the surface during the warm season, and the late-afternoon peak of diurnal precipitation variation can be explained by this fact (e.g., Wallace, 1975; Yu et al., 2007b, Yu et al., 2010; Zheng et al., 2008). Convective precipitation occurring at nighttime or in the morning is often associated with factors like the diurnal variation of local thermal circulation forced by complex terrain or the distribution of land and waters, nocturnal boundary layer friction, low-level jets, persistent MCSs, and inertia-gravity waves (e.g., Wallace, 1975; Zeng et al., 1994; Mori et al., 2004; Zheng et al., 2008; Chen et al., 2010, Chen et al., 2013, Chen et al., 2014, Chen et al., 2016, Chen et al., 2017; Fujita et al., 2010; Zheng and Chen, 2013; Du and Rotunno, 2018). Some studies (Yang and Slingo, 2001; Mori et al., 2004; Fujita et al., 2010) have proposed that gravity waves and cold gravity currents produced by existing MCSs, the self-replicating mechanism of MCSs, and background wind flows, are also influential factors that contribute to diurnal variations and propagations of precipitation. These factors listed above cause the bimodal feature of diurnal variation at each level over central and eastern China.

In regions A, B, and I, the nocturnal stratiform precipitation peak can be attributed to either the instability caused by nocturnal radiative cooling at the cloud top (Lin et al., 2000) or local thermal circulations like land-sea breezes (Chen et al., 2016; Zhu et al., 2017; Du and Rotunno, 2018). However, in regions C, D, and G, the nocturnal convective precipitation peak is often related to large-scale background circulation, low-level jets, interactions with existing MCSs, and local thermal circulation like mountain-plain solenoids (Luo et al., 2014; Wu and Luo, 2016; Tian et al., 2018; Zhang et al., 2018). In addition to the above factors, extreme precipitation occurrences are usually associated with favorable environmental conditions, especially an extremely abundant water vapor supply with total precipitable water exceeding 60 mm (Tian et al., 2015, 2017, 2018). A typical extreme precipitation case with total precipitable water up to 60 mm occurred on 7 May 2017 in Guangzhou (Tian et al., 2018). Although the present study reveals the diurnal variation features of extreme precipitation, the diurnal variation in extreme precipitation accumulated over a longer period still remains unclear. Thereby, diurnal variations in persistent extreme precipitation at various levels need to be explored in the future.

In Southwest China (regions E, F, and H), nocturnal peaks and propagations of precipitation can be explained by local thermal circulations such as mountain-plain solenoids (Yu et al., 2007b; Bao et al., 2011; Chen et al., 2013; Zhang et al., 2018). Nevertheless, the physical mechanisms for stratiform precipitation between midnight and early morning are complicated (Yu et al., 2007a), which might also be related to radiative cooling at the cloud top and decaying MCSs (e.g., Wallace, 1975; Houze, 1997; Lin et al., 2000; Yu et al., 2010). The fact that peaks of convective precipitation occur earlier than those of stratiform precipitation during the nighttime is possibly related to the fact that stratiform precipitation is generated by decaying MCSs that have already produced heavy convective precipitation in advance (Houze, 1997; Yu et al., 2010), but this should not be the only mechanism because the occurrence frequency of convective precipitation is far less than that of stratiform precipitation (Fig. 4).

Over the coastal regions, such as regions D-L, D-S, I-S (Fig. 6), 35^°-40^°N along 120^°E (the coastal region of the Bohai Sea; Fig. 8b), and 106^°-115^°E along 22^°N (the coastal region of South China; Fig. 9a), more precipitation occurs in the early morning. This fact can be associated not only with factors such as large-scale circulations, terrain, near-surface friction, and inertia-gravity waves (Zheng and Chen, 2013; Chen et al., 2014, Chen et al., 2016, Chen et al., 2017; Jiang et al., 2017; Du and Rotunno, 2018), but also with convergence caused by sea breezes from different directions or by interaction between large-scale circulation and onshore flow (Pielke, 1974; Zhu et al., 2017; Du and Rotunno, 2018).

More importantly, although many studies have investigated the mechanisms for diurnal precipitation variations at various regions, the detailed mechanisms for the variation features shown in the present study, such as the relationship between the variations and local thermal circulations in Southwest China (regions F and H), in the middle reaches of the Yangtze River and in the coastal regions like regions D-L, D-S, and I-S, are very complicated and still need further investigation.

On the basis of previous studies, the present study further analyzes the diurnal variations and regional differences of hourly precipitation at various levels that represent different types of precipitation based on no less than 30 years of observations collected at 2061 national weather stations in China during May-September. The major conclusions can be summarized as follows:

(1) As a whole, diurnal variations in amount and frequency of total, stratiform, convective, and extreme precipitation, over the whole of central and eastern China, all display a bimodal feature, with the primary peak occurring in the late afternoon and the secondary peak occurring between midnight and early morning.

(2) Normalized diurnal variation amplitudes of convective precipitation amount and frequency are generally larger than those of total precipitation and increase with an increasing threshold, and convective precipitation determines the variation of total precipitation intensity. In general, the diurnal variations in precipitation frequencies at the levels ≥ 0.5 mm h^-1 and ≥ 1 mm h^-1 are much more insignificant than those at other levels. The diurnal variation in precipitation at the level ≥ 10 mm h^-1 can represent well that of convective precipitation in China, while the diurnal variation at the level ≥ 5 mm h^-1 can represent well that of convective precipitation in northern China. This indicates that stratiform precipitation is on average heavier over southern China than over northern China. The normalized diurnal amplitudes of amount and frequency of hourly precipitation ≥ 50 mm are the most significant among all levels. Overall, the extreme precipitation intensity at the level ≥ 50 mm h^-1 at nighttime is slightly larger.

(3) Based on the primary peak time periods of precipitation frequency at various levels, the regions can be classified into three main categories: regions with a late-afternoon primary peak (regions B and I), regions with a nocturnal primary peak (regions E, F, and H), and regions with a time-shifting primary peak (regions A, C, D, and G). However, over some coastal regions of China, such as regions D-S, D-L, I-S, 35^°-40^°N along 120^°E, and 106^°-110^°E along 22^°N, the diurnal precipitation variations differ from those over their corresponding larger regions.

(4) For various regions, diurnal variations in amount and frequency of total and stratiform precipitation mainly feature a bimodal pattern, while the variations in convective precipitation generally show a late-afternoon primary peak in most regions. Nevertheless, over regions C, D, F, and H, convective precipitation frequency also displays an obvious secondary peak. In contrast, both convective and stratiform precipitation over region E exhibit a unimodal pattern, with the peak appearing after midnight. Furthermore, in Southwest China, with a nocturnal primary peak, and in the western North China Plain, all the peaks of convective precipitation at nighttime occur ahead of their stratiform precipitation counterparts, and the peak occurs earlier as the intensity of convective precipitation increases.

(5) Diurnal precipitation variations along some typical meridians and parallels present multiple patterns and propagations of precipitation. Propagation features of convective precipitation are different from those of total precipitation over 31^°-37^°N along 115^°E, and over 111^°-115^°E along 30^°N.