1 Introduction

The temporal and spatial variation of precipitation is a direct reflection of change in the climate system on the global and regional scales. Such variation is also closely associated with the hydrological cycle, again both globally and regionally. Considerably more or less regional precipitation usually causes flooding or drought, respectively. Within precipitating cloud, the latent heat released through phase change accounts for approximately three-quarters of the energy needed for atmospheric circulation, so such cloud plays a role in the process of energy balance in the earth–atmosphere system. In addition, precipitation has a substantial effect on the dilution of sea water salinity near the sea surface, as well as the wet deposition of air pollutants and aerosols in the lower atmosphere. As an important indicator of global and regional climate change, precipitation has become an increasingly critical diagnosis factor of climate change (Wang et al., 1987,2002;Wang, 1994) and a very direct parameter for the evaluation of weather and climate models. The methods of precipitation observation or measurements are of strategic significance in monitoring and studying regional or global climate changes.

The methods of precipitation observation or measurements are varied. Traditional observation by rain gauges is highly precise at the observational points, contrary to their inhomogeneous distribution spatially. In particular, the density of rain gauges is always sparse in mountainous regions, where torrential rainfall occurs frequently, and thus it is difficult to capture the large-scale precipitation distribution accurately by using rain gauges. Conversely, ground-based precipitation radar can usually detect rainfall over large-scale ranges of around 300 km. However, it is impossible to use ground-based radar measurements over extensive ocean areas or places difficult to reach for human beings. Besides, some serious biases can be caused by the influences of echo weakening, echo overshoot, and fake echo by topography. With the development of satellite remote-sensing techniques, satellite-based methods of precipitation measurement have become widely used in the monitoring of precipitation on larger scales and at finer resolution. A typical example of such progress is the Tropical Rainfall Measuring Mission (TRMM) satellite, jointly overseen by NASA and JAXA (Japan Aerospace Exploration Agency), which has enabled us to measure clouds and precipitation in the tropics and subtropics since its launch in 1997 (Kummerow et al., 1998). There are five sensors onboard the TRMM platform, including the Precipitation Radar (PR), CERES (Clouds and the Earth's Radiant Energy System), the Lightning Imaging Sensor (LIS), the TRMM Microwave Imager (TMI), and the Visible and Infrared Scanner (VIRS). As such, TRMM is a highly integrated meteorological satellite capable of active and passive observation of clouds and precipitation, although drizzle or light rain is difficult to detect by using PR with its wavelength of 2.2 cm. Moreover, the PR-derived near-surface rain rate also shows a suspiciously extreme rainfall profile, which is caused by contamination from ground or sea clutter (Hamada and Takayabu, 2014). Therefore, the accuracy of PR-derived light rain or extreme rainfall should be validated by rain gauges or ground-based radar observations. In spite of this, the long time series of PR data means it is highly applicable in studies of precipitation characteristics; its uniform measurements over land and ocean in the tropics and subtropics help to elucidate the nature of cloud and precipitation in this regions, which remains an open question in meteorological and climate science.

TRMM PR has been in operation for almost 17 years, resulting in huge quantities of data having been obtained. Great achievements with respect to clouds and precipitation in the tropics and subtropics have been acquired at home and abroad on the basis of TRMM PR datasets. Such researches by using TRMM PR measurements has focused mainly on deriving and documenting statistical characteristics of clouds, precipitation, and lightning, including: (1) climatic diagnosis of the relationship between abnormal changes of clouds, precipitation and lightning, and atmospheric low-frequency oscillation or El Niño (Anyamba et al., 2000;Cecil et al., 2005;Li et al., 2005,2011;Masunaga et al., 2006;Lin et al., 2013); (2) the diurnal cycle of clouds and precipitation (Fujinami et al., 2005;Ichikawa and Yasunari, 2006;Sanderson et al., 2006;Yang and Smith, 2006;Yamamoto et al., 2008;Chen et al., 2009;Liu P. and Fu, 2010;Yu et al., 2010;Mao and Wu, 2012); and (3) the characteristics of precipitation for regional weather and climate (Petersen et al., 2002;Kodama and Yamada, 2005;Liu and Fu, 2007;Liu Q. and Fu, 2010;Yuan and Qie, 2010;Du et al., 2011;Xu, 2013;Song and Sohn, 2015).

This review summarizes research progress since the turn of the century on the analysis of the characteristics of East Asian summer precipitation from a climatological perspective, based on TRMM PR measurements. We focus mainly on the climatological characteristics of precipitation in terms of its horizontal distribution, vertical structure, diurnal cycle, and recent long-terms trends.

2 Climatology of the three-dimensional structure of summer precipitation 2.1 Horizontal distribution

To some extent, the horizontal distribution of precipitation parameters (e.g., rain amount, rain rate, rain type, rain frequency, and so on) can reflect the nature of precipitating cloud systems and their states (Hobbs, 1989;Zipser and Lutz, 1994;Fu et al., 2007). Based on different microphysical processes, precipitation is mainly categorized into convective precipitation and stratiform precipitation under the classification of TRMM PR (Simpson et al., 1988;Iguchi et al., 2000). For East Asia, studies have revealed that both convective and stratiform precipitation is closely associated with southerly or southwesterly winds at 850 hPa, and also modulated by the subtropical ridge of the western Pacific (Fu et al., 2008;Lu et al., 2016). In the climatological mean pattern (Figs. 1a and 1b), the horizontal distributions of convective and stratiform rain rates are generally similar, but the convective rain rate (5.0–9.0 mm h^–1) is more than three times that of the stratiform rain rate (1.2–4.0 mm h^–1;Fu and Liu, 2003;Fu et al., 2003;Yang, 2015;Lu et al., 2016). However, the frequency of stratiform precipitation is higher than that of convective precipitation over most of East Asia, which means stratiform precipitation is the main rain type in this region.

In order to comprehensively reveal the spatial patterns of both convective and stratiform precipitation, previous studies (Fu et al., 2003;Lu et al., 2016) have employed the area fraction (AF) of precipitation to quantitatively assess the ratio of each precipitation type (stratiform/convective), and the contribution fraction (CF) of precipitation to represent the rainfall contribution of each precipitation type to total rainfall. The results have shown that stratiform precipitation is the main form of rainfall in summer in East Asia, with the AF as high as 80% (Fig. 1c). In detail, higher AFs for stratiform precipitation (> 70%) are mainly located in the northern part of East Asia (north of 25°N), whereas the higher AFs for convective precipitation (> 20%) mainly appear in the southern part (south of 25°N;Figs. 1c and 1d) (Fu et al., 2007;Lu et al., 2016). However, because the conditional rain rate of convective precipitation is nearly four times larger than that of stratiform precipitation, convective, and stratiform rainfall contribute almost equally to total precipitation over most regions in East Asia (Figs. 1e and 1f) (Fu et al., 2003,2007;Liu P. and Fu, 2010;Lu et al., 2016). For most areas of the Yangtze-Huaihe River valley in the mei-yu season, the CFs of stratiform precipitation are more than 50%, whereas the CFs of convective precipitation mainly range from 20% to 40%. In the middle reaches of the Yangtze River and the upper reaches of the Huaihe River, the CFs of stratiform precipitation are less than those of convective precipitation (Yang et al., 2014).

Previous work has also indicated the distribution of storm rain in southern China, based on 10-yr measurements by TRMM PR. It has been found that the storm rain frequency is more than 1.2% in southern China, with the maximum even exceeding 1.8% (Fig. 2a). Meanwhile, the frequency of heavy/extra storm rain exceeds 0.060%/0.015% in southern China, which appears in a band-like distribution with northeast–southwest orientation (Figs. 2c and 2e). Statistically, the rainfall contributions of storm, heavy storm, and extra storm rain to total rainfall are over 25%, 5%, and from 1% to 3%, respectively (Figs. 2b,2d,2f) (Fu et al., 2011).

Additionally, a number of studies have compared the rain rates measured by the PR with that by rain gauge observations on the climatological scale. For example,Liu et al. (2010) compared PR measurements and rain gauge observations from 1998 to 2005, and their results indicated that the spatial patterns of the summer mean rain rate calculated by PR and rain gauge datasets are nearly consistent, as shown in Fig. 3. However, it is obvious that values of both rain rates are clearly different in southern China. As pointed out in their paper, the mean rain rate obtained by rain gauge observations is relatively higher than that by the PR. Also, differences between the PR and rain gauge observations in extreme values and their coverage are also clear in South China, mainly determined by the spatial density of the rain gauges. Essentially, such differences are generated from the different manner of observations between the PR and rain gauges. The manner of the PR's measurements is an instantaneous scan over the coverage of precipitating clouds, whereas rain gauges take accumulative measurements for a period at one spatial point. The light rain missed by TRMM PR measurements, which also causes a certain discrepancy in the rain rate when compared with that obtained by rain gauges, may be detected by the GPM (Global Precipitation Measurement) network of satellites' Ka-band radar, which has a shorter wavelength than PR. However, the inconsistency in the rain rate produced by different manners of observation is not easily overcome.

It is well known that in the middle of summer precipitation usually manifests in a northeast–southwest oriented band, widely known as the mei-yu front, over East Asia, from South China, via central eastern China and the East China Sea, to the South Korean Peninsula and South Japan. The front features relatively large quantities of rainfall, as observed by rain gauges over land. However, based on TRMM PR data, studies have indicated that summer precipitation is distributed around the western Pacific subtropical high, from its western to its northwestern and northern edge. The precipitation along the mei-yu front is one part of the surrounding precipitation (Fu et al., 2008). Studies have also found that the daily mean precipitation in summer over ocean areas is larger than that over land in the midlatitudes of East Asia, as measured by PR. The relatively larger values of greater than 8 mm day^–1 occur from the western Pacific Ocean to South Japan (Fu et al., 2008). One possible reason may be that clouds over ocean areas can easily obtain much more water vapor from the surface of the ocean. The above results highlight the role of PR measurements over the ocean, where rain gauges are unable to help.

It is noteworthy that recent studies have paid particular attention to studying the relationship between precipitation features and geomorphic characteristics in East Asia. Considering East Asian monsoonal precipitation progresses in a distinct seasonal manner (Wang and Lin, 2002),Xu (2013) pointed out that the most intense storms usually occur in the South China region, followed by the lowland, plateau, foothill, and ocean regions.Song and Sohn (2015) found that the seasonal march of the East Asian and western North Pacific summer monsoonal precipitation also displays a distinct stepwise northward and eastward advance from June to August. Their results also showed that most heavy rain with high storm height and abundant ice water, i.e., cold storms, occurs over inland China due to the atmosphere there under convectively unstable conditions, whereas most warm storms associated with a lower storm height and lower ice water content appear over the ocean. Studies have also revealed different meteorological environmental conditions and atmospheric dynamic/thermodynamic environments in different East Asian regions during monsoon season, e.g., the convective available potential energy (CAPE), height of neutral buoyancy, and vertical wind shear (Xu, 2013;Song and Sohn, 2015).

Generally, although stratiform precipitation and convective precipitation occur simultaneously in East Asia, and the rain rate of the former is smaller than that of the latter, stratiform precipitation has a higher frequency and so is still the dominant type of precipitation. On the contrary, although the rain rate of convective precipitation is larger than that of stratiform precipitation, the lower frequency of the former means both stratiform and convective precipitation contribute almost equally to total rainfall over East Asia. The features of precipitation over East Asia are modulated by environmental conditions, e.g., atmospheric stability, humid condition, and local topography. Note that Figs. 2 and 3 only cover the mainland of China, mainly representing the precipitation characteristics over the land areas of East Asia. In addition, the studies reviewed here focused on the summer seasonal mean state. It should be noted that month-to-month variations are still identifiable, which is a manifestation of the entire seasonal evolution of the East Asian monsoon and its associated precipitation, as indicated by Wang and Lin (2002) and Song and Sohn (2015).

2.2 Vertical structure

The vertical profile of rain rate and storm-top height are the most important parameters for precipitating clouds in terms of the thermal and dynamic conditions associated with microphysical processes (Hobbs, 1989;Zipser and Lutz, 1994;Fu et al., 2007;Xu, 2013;Song and Sohn, 2015), directly reflecting how well a storm develops vertically. These profiles and tops can be detected and retrieved from radar echoes measured by the PR.

Previous studies have shown that there are obvious differences among the vertical rain profiles of convective, stratiform, and other types of precipitation (Fu and Liu, 2001). Using PR rain profiles with empirical orthogonal function (EOF) analysis, it was found that the first leading principal component of the EOFs can explain greater than 85%/80% of total variances of all vertical profiles for stratiform/convective precipitation over East Asia, when given a near-surface rainfall rate (Fu et al., 2003,2007). Therefore, it can be deduced that the mean vertical profile presents a very good representation for studying the rain profiles over East Asia, because statistical physical laws can be extracted from the vertical structure of the precipitation cloud, determined by the same existing microphysical process (Fu and Liu, 2001;Li et al., 2015). For example, a rain rate increase (decrease) towards the surface in the precipitation profile indicates the precipitation particles are increasing (breaking or evaporating) during descent (Hobbs, 1989;Fu et al., 2007,2008). According to the slopes of the precipitation profile, the mean profiles for convective and stratiform precipitation can be divided into four layers and three layers, respectively, which represent the microphysics of precipitation (Fu et al., 2003,2007,2008;Yang, 2015).

For summer convective precipitation profiles in three typical regions of East Asia (continental land, the East China Sea, and the South China Sea;Figs. 4a–c), all of them present four regimes, from lower to upper levels, as follows: (1) a layer of evaporation from the surface to an altitude of approximately 4 km (3 km) over land (ocean); (2) a layer of coalescence from the layer of evaporation to the freezing layer (approximately 5 km over land and approximately 4.5 km over ocean); (3) a mixed layer of ice and water from the layer of coalescence to 7 km over land or 6.5 km over ocean; and (4) an ice layer above the mixed layer of ice and water. In contrast, because of the cold underlying surface, the layer of evaporation disappears in winter convective precipitation profiles, which only has three regimes and their heights are lower than those in summer (Figs. 4d–f).

As for stratiform precipitation, its profiles in summer for the three typical regions in East Asia, shown in Figs. 4g–i, indicate features in three regimes, from lower to upper levels, as follows: (1) a uniform layer below the freezing layer in which the rain rate stays approximately the same with the change in altitude; (2) a mixed layer of ice and water from the freezing layer to an altitude of 6 km (the same over land and ocean); and (3) an ice layer above the mixed layer of ice and water. Similarly, almost the same three regimes in are apparent in the winter stratiform precipitation profiles, apart from a lower freezing level height (Figs. 4j–l).

In conclusion, the main difference between convective and stratiform precipitation profiles is the lack of evaporation and the raindrop coalescence during raindrop descent for stratiform precipitation, which is probably due to the different microphysical and dynamic processes. Meanwhile, the land–ocean difference in rain profiles is mainly induced by the difference in updraft and thermal processes in precipitation clouds (Zipser and Lutz, 1994). It is logical to speculate that the reasons for these differences could derive from various factors, including different air mass properties, SST, and atmospheric stability, which are regionally dependent. There is a definite need for follow-up studies that help us to fully understand the regional differences in precipitation profiles.

Xu (2013) recently reported different vertical structures of rain clouds occurring between mainland China and the East China Sea by using TRMM PR measurements, e.g., more vigorous convection is found in the eastern part of mainland China.Song and Sohn (2015) further showed a huge contrast between continental convection over mainland China and oceanic convection over Korea, Japan, and surrounding oceans. The west-east differences in the characteristics of rain over East Asia mainly originate from the different ambient meteorological conditions, such as the CAPE, height of neutral buoyancy, total water vapor, and vertical wind shear, in different regions (Xu, 2013;Song and Sohn, 2015). It is interesting that a striking feature is the maximum variance appearing at the freezing level regardless of rain type, season, and location. The variances become larger when the surface rain rates increase (Fu et al., 2003). All these aspects imply a very complicated rain formation process near the freezing layer, which should have referential meaning for simulations of hydrological processes in numerical weather and climate models.

The storm-top altitude is defined as the maximum height that precipitation particles can arrive in precipitating cloud, which is related to the updraft strength and atmospheric stability (Chen et al., 2016). This top altitude, as detected by radar, is related to radar wavelength: the longer the wavelength, the lower the top altitude. Therefore, the storm-top altitude here only refers to that measured by the PR. For stratiform precipitation, most storm-top altitudes vary from 5 to 7 km, regardless of whether over land or ocean (Fig. 5a) (Fu et al., 2012;Yang, 2015;Chen et al., 2016). Meanwhile, for convective precipitation, with thicker and higher clouds (Fu et al., 2003,2007), more than 70% of convective storm tops range from 5 to 10 km over the eastern plain of mainland China, while they are between 5 and 9 km in other regions (Fig. 5b) (Fu et al., 2012;Yang, 2015;Chen et al., 2016). Generally, the mean convective storm top is higher over land than over ocean, e.g., the mean height of the storm top for a surface rain rate of 30 mm h^–1 in summer can reach approximately 16 km over the East Asian continent, whereas the value over ocean is only around 14 km (Fu et al., 2007). It can be concluded, therefore, that convective precipitation over land is easily forced by the land surface, whereas stratiform precipitation is not. Moreover, more than 40% of deep convective precipitation with a mean storm top of greater than 10 km appears mainly over the eastern plain of mainland China and Southwest China (Fu et al., 2012), whereas most heavy stratiform precipitation tops are less than 8.5 km. It can also be deduced that the differences in regional storm tops are mainly modulated by different ambient meteorological conditions (Xu, 2013;Song and Sohn, 2015). However, overall, more observations are needed to further study these ideas.

Based on TRMM PR datasets, the relationship between rain intensity and storm-top altitude in summer over East Asia has been revealed. Statistics show that the mean altitude of storm top for both convective and stratiform precipitation increase with the increment in the mean surface rain rate, and relationship between the altitude of storm top and rain rate show remarkable quadratical function (Fu et al., 2012;Chen et al., 2016). Studies have also exposed that the mean thickness of the anvil is about 3–4 km, with its bottom located at an altitude of 6 km and its top at 10–12 km, over East Asia in summer (Fu et al., 2010). The above results on the relationship between rain intensity and storm-top altitude can potentially be applied in parameterizing precipitation in numerical weather or climate models.

At present, achievements remain somewhat limited. For example, statistically, how the vertical structure of precipitation changes during the development of a precipitating system; the extent of the differences in these changes between regions in East Asia; and the connection between these changes and the environmental parameters of surface temperature, total water vapor, atmospheric stability, available potential energy, vertical wind shear, and so on, in regions of East Asia. To address these problems, more observational experiments and model simulations are needed in the future.

3 Diurnal variation of summer precipitation

Diurnal variation is the most fundamental mode of precipitation variability, due to the weather and climate system responding to solar radiation forcing on the rotation and revolution of earth's surface. The diurnal variation of summer precipitation over East Asia has been widely investigated using the TRMM PR datasets (Liu P. and Fu, 2010;Yu et al., 2010;Fu et al., 2012;Mao and Wu, 2012).

Using a relatively short TRMM dataset (1998–2003),Yamamoto et al. (2008) studied the peak time of precipitation over the tropics and midlatitudes from the coldest minimum brightness temperature and the maximum rain rate. Based on a 10-yr set of TRMM PR measurements,Fu et al. (2012) found that the distributions of the diurnal peak for parameters (rain frequency, rain rate, and storm top) of both stratiform and convective precipitation present a consistent variation trend in summer over East Asia, as shown in Fig. 6. For convective precipitation, its peak local time for precipitation frequency, intensity and storm-top altitude occurs from afternoon to evening over the Tibetan Plateau, most parts of southern China, the Indochina Peninsula, and most parts of India. However, over the Sichuan basin, the steep areas in the south of the Himalaya, and the valley area located on the southwest side of the Hengduan Mountains and southeastern Tibetan Plateau, the peak local time for these three parameters appears mainly from midnight to early morning. Contrary to the clear diurnal peak local time for convective precipitation over land, its peak local time (LT) over ocean seems to be at noon (1100–1400 LT) or in the evening (1900–2200 LT), except over islands where the peak local time for convective precipitation shows up in the afternoon (1500–1800 LT). For stratiform precipitation, basically, the distribution of its peak local time for these three parameters is similar to that of convective precipitation. However, when comparing the diurnal peak of local time for convective precipitation with that for stratiform precipitation, it is clear that the former is earlier than the latter. For example, the diurnal peak of convective precipitation occurs in the afternoon in the southeast highlands of China, whereas that of stratiform precipitation appears in the evening, which implies that stratiform precipitation very likely evolves from convective precipitation in the region. In the Sichuan basin, from the south of the Tibetan Plateau via the southeastern Tibetan Plateau to northeastern Indochina, more obvious nighttime rain occurs for stratiform precipitation than convective precipitation, as shown in Fig. 6. Over ocean, usually, the area with the same diurnal peak of local time is larger for stratiform precipitation than convective precipitation, possibly because of the larger coverage of stratiform precipitation. Therefore, the diurnal peak of local time for stratiform precipitation shows more integrity over ocean.

To better comprehend the regional differences in the diurnal variation of precipitation over land and ocean in East Asia,Fu et al. (2012) plotted the diurnal cycle of convective precipitation frequency, intensity, and rain-top altitude in eight regions in East Asia and South Asia: the Indian subcontinent, the Bay of Bengal, the South China Sea, the warm pool of the western Pacific Ocean, the East Plain of China, the southeast highlands of China, the Tibetan Plateau, and the Sichuan basin (Figs. 7 and 8). The results indicate that the local time peaks of three parameters for summer convective precipitation over land, such as over the Indian subcontinent, the Tibetan Plateau, and East and South China, mainly occur from afternoon to evening (Fig. 7)—except in the Sichuan basin, where the local time peaks of the three parameters appear in the early morning, which is the so-called “Bashan nocturne.” Over ocean, the amplitude of the diurnal cycle for these parameters is obviously smaller than that over land. The above results indicate that the underlying surface type plays a very important role in the diurnal cycle of convective precipitation. But more specifically, how does the underlying surface—including the parameters of surface temperature, water vapor, and atmospheric stability—quantitatively impact on the diurnal cycle of convective precipitation? The differences among these peaks between island and ocean mainly appear at noon and the period from noon (1100–1400 LT) to afternoon (1500–1800 LT), respectively.

For stratiform precipitation, as shown in Fig. 8, the diurnal cycle pattern of stratiform precipitation frequency, intensity, and rain-top altitude in eight regions is similar to that of convective precipitation. However, the amplitude of the diurnal cycle for the three parameters for stratiform precipitation is smaller than that for convective precipitation. Besides, the local time of the peaks for the three parameters for stratiform precipitation seems to lag behind that for convective precipitation by 1–2 hours. This feature is clear in the Sichuan basin, Tibetan Plateau, and East and South China. The lag of stratiform precipitation behind convective precipitation over land in East Asia, also pointed out by Yamamoto et al. (2008), suggests that stratiform precipitation possibly evolves from convective precipitation.

Whilst the topography of East Asia is a very important aspect influencing the diurnal variation and phase propagation of precipitation in summer over this region (Liu P. and Fu, 2010;Fu et al., 2012;Mao and Wu, 2012), the activity of atmospheric circulation must also be considered. For example, in central eastern China, the afternoon peak of convective rainfall is usually forced by local thermal forcing, such as the land surface being heated by solar radiation; whereas, the peak from evening to early morning of stratiform rainfall, evolved from convective precipitation, is still modulated by large-scale circulations, such as the diurnal variation of low-level wind circulation induced by the interaction of low-level atmospheric circulation and topography (Chen et al., 2010;Yu et al., 2014). The above results show some similarity to those of Yamamoto et al. (2008).

Compared with the diurnal variation observed by rain gauges (Yu et al., 2007a,b;Zhou et al., 2008) or by TRMM TMI (Chen et al., 2009;Mao and Wu, 2012), the advantage in studying precipitation diurnal variation observed by TRMM PR is that PR can also reveal the diurnal variation of precipitation in terms of vertical structure and rain type (Yu et al., 2010;Fu et al., 2012), which is useful for evaluating the cloud and precipitation simulated by models. For instance,Yu et al. (2010) pointed out that the maximum stratiform rain rate and the highest profile of stratiform precipitation appear in the late-night (late-afternoon) over southwestern (southeastern) China, whereas most of the short-duration stratiform rain rate tends to exhibit late-afternoon peaks over southern China. For convective precipitation, the maximum rain rate and the highest profile appear in the late afternoon over most of southern contiguous China, whereas the long-duration convective rain rate presents late-night peaks over southwestern China.

In addition, it should be mentioned that the low temporal resolution of PR-observed sampling is a concern when studying the diurnal variation of precipitation, or even other climatological characteristics, as compared with the higher temporal resolution of samplings observed by rain gauges. Many short-term convective precipitation events might not be captured by TRMM PR because of its low-resolution spatiotemporal sampling. However, considering the random occurrence of convective events in both space and time, statistically, this large amount of data should be able to capture the climatological features of convective precipitation events (Yama-moto et al., 2008;Fu et al., 2012).

Meanwhile, from the perspective of satellite remote sensing, there are many aspects relating to studies on the diurnal variation of precipitation that need to be considered. For example, instruments onboard satellites detect different targets, such as the precipitation profiles measured by radar, the optical spectrum scanned to establish cloud tops, and the droplets and vapor that can be probed by passive microwave imagers. Therefore, the diurnal variation of cloud top, near-surface rain rate, storm top, and precipitable water, as well as the relationships among them, is worthy of study.Yamamoto et al. (2008) preliminarily identified relationships between TRMM sensor signatures and the development of a convective rain system. Despite the above work, this possible dominant mechanism of East Asian precipitation needs be further considered in model simulations, and also validated by more observations, in future studies.

4 Recent trends of summer precipitation

In summer, East Asia and its surrounding areas are directly dominated by the monsoon, which varies strongly in time and space. For instance, the interannual variation of the summer monsoon has a significant impact on local atmospheric parameters including temperature and precipitation. The temporal and spatial changes of these parameters, in turn, cause uncertainty around the occurrence of extreme weather such as flooding, drought, and heat waves (Ding, 2007;Ding et al., 2009,2013). Previous studies have revealed the trends of precipitation in China by using rain gauge data. There is a clear increase in the boreal-summer precipitation amount over the Yangtze River basin, but a decrease in northern China (Hu et al., 2003;Yang and Lau, 2004;Ding et al., 2007,2008). Meanwhile, rapid industrialization in China has unavoidably led to a dramatic increase in aerosol concentrations (Tie et al., 2006), similar to what developed countries once experienced (Mayer, 1999). The impact of aerosols on cloud and precipitation—the so-called aerosol indirect effect (Twomey, 1974;Albrecht, 1989)—is one of the most challenging problems in climate research. The effects on precipitation in East Asia of the weakening East Asian summer monsoon (EASM), along with those of human activities related to urbanization and industrialization, is a key focus for atmospheric scientists. Specifically, we are seeking to address two important questions: (1) What are the relationships between the interannual variability of convective or stratiform precipitation and the EASM? (2) What is responsible for the precipitation changes—natural variability or human activities?

In terms of the first question,Lu et al. (2016) investigated the interannual variation patterns of convective and stratiform rain rates (RRs), area fractions (AFs), and contribution fractions (CFs), in summer over East Asia. They found that the spatial patterns of intensities for the interannual variability (i.e., standard deviation/coefficient of variation) of RRs/AFs/CFs are similar between stratiform and convective precipitation (Fig. 9), and concluded that such interannual variation of precipitation is mainly influenced by the interannual variation of the EASM. Notably, only the coefficient of variation for the convective AF is much greater than that for stratiform precipitation; the coefficients of variation for both stratiform and convective RR/CF are similar in most areas of East Asia. Through further analysis of the variational relationships among stratiform and convective precipitation, the EASM, northwestern Pacific subtropical high, East Asian westerly jet, and tropical SST anomalies on an interannual scale, it was deduced that the intensity of the summer monsoon in the lower troposphere mainly influences the interannual variations of stratiform/convective RR and atmospheric stability, which further modulates the interannual variations of stratiform/convective AF and CF. Despite the above work, this possible dominant mechanism of East Asian precipitation needs be further considered in model simulations, and also validated by more observations, in future studies.

Regarding the second question, the recent investigation by Fu et al. (2016) revealed that the rain frequency (RF) for both convective and stratiform precipitation increases in the majority of regions in southern East China [SEC; (26°–30°N, 113°–122°E)] but decreases in the northwestern part of northern East China [NEC; (30°–35°N, 113°–122°E)], as shown in Fig. 10. The decreasing rate of RF for stratiform precipitation in NEC is twice as much as that for convective precipitation, while the increase of convective precipitation in SEC is more evident than for stratiform precipitation. The RR exhibits a decreasing trend in most portions of East China for both convective and stratiform precipitation. These trends for both convective and stratiform precipitation are possibly affected by the different levels of aerosols in SEC and NEC. To understand how natural factors impact these trends, two important components—the precipitable water (PW) and water vapor transport (WVT), represented by the column-integrated water vapor and vertically integrated atmospheric water vapor transport from adjacent regions, respectively—were calculated in the East Asian monsoon system. The results, as shown in Fig. 11, revealed that, in NEC, with an increasing trend of aerosol optical depth (AOD), the decreasing trends of both WVT and CAPE are the same as that of rain rate and rain frequency for both convective precipitation and stratiform precipitation. Meanwhile, in SEC, with a clear decreasing trend of AOD, the increasing trends of rain frequency for both convective precipitation and stratiform precipitation are the same as those of PW and WVT. The above results suggest that precipitation occurs different performances under different human and atmospheric environment conditions, such as in NEC or SEC.

On the other hand, a previous study found evidence that aerosols suppress precipitation by increasing the atmospheric stability (Zhao et al., 2006). Aerosols can absorb sunlight, heating the air, and then increasing the atmospheric stability, leading to a reduction in precipitation, i.e., in a positive feedback cycle (Zhao et al., 2006;Rosenfeld et al., 2008). In fact, the aerosol conditions are different between NEC and SEC in East China making it an ideal region to examine the indirect effect of aerosols on convective and stratiform precipitation. The mean AOD [data taken from MODIS (Chu et al., 2002)] ranges from 0.3 to 0.9 in summer in East China. Heavy aerosol loadings with AOD exceeding 0.6 are found in NEC, whereas it is relatively clean in SEC (Fu et al., 2016). Their results showed that regions under heavy aerosol conditions (i.e., NEC) exhibit downward trends for RR and RF for both rain types, especially stratiform precipitation (Fig. 10). The area-averaged CAPE exhibits a relatively clear positive trend in SEC and negative trend in NEC, suggesting the atmosphere has become more stable in NEC and more unstable in SEC over the last 11 years (Fig. 11). The above results are consistent with evidence of the suppression of precipitation by aerosols through increasing the atmospheric stability (Zhao et al., 2006). Therefore, heavy aerosol loadings is one possible mechanism through which precipitation in NEC is reduced. However, in SEC, it is also reasonable to consider that the increasing trend of stratiform rain frequency in the region induces more wet scavenging, thus reducing AOD. More detailed studies are needed to confirm these inferences.

Generally, the atmospheric stratification and moisture transportation in the East Asian monsoon area are controlled by the summer monsoon (which is an element of natural variability), and then the rain rate and frequency of stratiform precipitation change and the total precipitation amount is affected. Besides, regional discrepancies exist in AOD loading, which can also change the atmospheric stability, such that the rain rate is affected by aerosols (which is an element of anthropogenic influence). Therefore, against the background of a weakening East Asian monsoon, the precipitation amount and its rain rate in most regions of East Asia are more affected by monsoon activity, whereas aerosols also have an influence on the rain rate and frequency of rain type over some high pollution regions (e.g., eastern China).

Thus far, the relationship between precipitation and aerosol loading is a bifunctional feedback problem tangled with many factors. As suggested by Zhao et al. (2006), there is a possible feedback cycle between aerosol and precipitation in eastern central China; that is, more aerosols tend to increase the atmospheric stability, leading to less precipitation, which causes less wet deposition of aerosols (more aerosols). On the other hand, aerosols are also affected by precipitation via wet scavenging, and this is also very important. Anyway, as mentioned above, the interactions among aerosols and precipitation is a complex question. Also of importance is that optical instruments onboard satellites cannot observe aerosols in cloudy sky, i.e., the AOD cannot be generated by MODIS during precipitation. Therefore, more ground-based instruments and modeling studies are needed to solve this question in the future.

5 Summary

From the perspective of climatology, this paper summarizes the characteristics of the three-dimensional structures of summer precipitation and their diurnal variation over East Asia. Meanwhile, from the perspective of climate change, this paper reviews the interannual variability of summer convective and stratiform precipitation and their correlation with large-scale circulation under the global-warming scenario and against the background of a weakening EASM, as well as the recent trends of summer convective and stratiform precipitation in central eastern China. Aside from research on the climatological diagnosis of precipitation by using TRMM PR, on the basis of TRMM multi-sensor observations there are still many other research fields worthy of attention, briefly summarized as follows: (1) Data from multiple instruments onboard TRMM have been archived and made available for more than 17 years. By matching and merging pixels of PR, VIRS, TMI, LIS, or clusters for rain and clouds, we can use the optical spectrum, passive microwave, and precipitation profiles of the merged data to reveal the characteristics of spectral and microwave signals (multichannel combined signals) of precipitating clouds, or to test the retrieved rain rate by infrared and microwave signals. (2) We can also use the above matched and merged data with a cloud parameter retrieval algorithm to study the relationship between cloud parameters and precipitation intensity. (3) By combining the merged precipitation data with reanalysis data or sounding data, we can analyze the atmospheric structure of temperature, moisture, and atmospheric stability in precipitating cloud, similar to a recent preliminary exploration (Xia and Fu, 2016).

When conducting these proposed studies, a number of challenges exist. It is still uncertain as to how to exploit detection datasets from different types of spacecraft, e.g., geostationary and polar-orbiting satellites, with reduced biases. For example, merging the data mentioned above with FY-4 series geostationary satellite data, or GPM or Himawari series data. It is also challenging to reveal the relationships between the intensity of precipitation and aerosols, because both variables cannot be measured at the same time by satellite sensors.

Although China's meteorological satellite program started relatively late, great progress has been made in recent years. However, we still need to learn from experience, as the old saying goes: “Stones form other hills may serve to polish the jade of this one” (advice from others may help one overcome one's own shortcomings). On a scientific level, we need to have a strategic vision and layout; and on an engineering level, we need to improve design and manufacturing technology. In pace with China's continual development, the status of our meteorological satellite program in terms of its international standing will improve. This has been proven by the successful launch and operation of the FY series of meteorological satellites (Zhang et al., 2009;Lu et al., 2011;Yang et al., 2012;Wang et al., 2014;Chen et al., 2016).

As a continuation of TRMM, the GPM network of satellites was launched on 28 February 2014, with a dual-frequency precipitation radar and microwave imager onboard, which will provide global observations of cloud and precipitation. Research on precipitation retrieval methods, ground correction, and the application of observational results are in full progress (Chandrasekar and Le, 2015;Liu and Zipser, 2015;Tang et al., 2016). China is also developing our first space-borne precipitation radar, and the operation of our own space-borne precipitation radar along with GPM will provide us with more detail on the characteristics of precipitation, which is critical for the verification and evaluation of weather and climate models, as well as studying the latent heat of precipitation.

Acknowledgments . This review is dedicated to commemorating the Chinese climatologist Shaowu Wang, who was one of the committee members for the doctoral dissertation oral defense of the first author on 31 August 1993. The comments from the reviewers and Dr. Xiangdong Zhang are appreciated.