Diurnal cycles of total rainfall, precipitation features, mesoscale convective systems (MCSs), deep convection, precipitation vertical structure, and lightning over the eastern Tibetan Plateau (TP) and eastward through China are investigated using 11 yr of Tropical Rainfall Measuring Mission (TRMM) measurements. Diurnal cycles of rainfall and precipitation features present apparent phase propagation eastward from the eastern TP for about 1000 km. The phase propagation is most evident during the pre-mei-yu and mei-yu seasons. However, it weakens with the northward progress of the East Asian monsoon and ceases in midsummer. During the pre-mei-yu season, diurnal cycles of storm population, total rainfall, deep convection, and lightning over the central and eastern TP foothills are in phase, peaking during the early morning. Another striking feature of the pre-mei-yu season is that nocturnal rainfall and MCSs prevail over the southeastern TP foothills following the deep convection and lightning maxima. These nocturnal peaks of deep convection and lightning over the foothills shift to afternoon after the onset of the monsoon, but nocturnal precipitation still dominates. Over the less mountainous region of eastern China, deep convection usually has an afternoon peak during the mei-yu whereas the rainfall maximum is at night. In midsummer, most parts of eastern China have strong afternoon peaks of deep convection, precipitation, and lightning, except in northeastern China where deep convection has an afternoon peak followed by a nocturnal precipitation peak.
The diurnal cycle of precipitation and convection is one of the most fundamental characteristics of a regional weather regime. It provides an important test for the validation of model physics in both weather and climate models (Lin et al. 2000; Trenberth et al. 2003; Dai et al. 1999; Dai and Trenberth 2004). Because of the complex terrain and evident seasonal change of weather regimes (Tao and Chen 1987; Chen 2004; Ding and Chan 2005; Xu et al. 2009), East Asia has been used frequently for investigating diurnal variations of rainfall and convective activities and their underlying physics (summarized in, e.g., Domros and Peng 1988; Zhao et al. 2005).
Over East Asia, in addition to the widespread afternoon rainfall peak, nocturnal rainfall peaks are often found in the valleys, foothills of high terrain, and over lakes and coastlines due to the low-level convergence by mountain–valley breezes and land–sea breezes (Ohsawa et al. 2001; Fujinami et al. 2005; Hirose and Nakamura 2005; Chen et al. 2005; Li et al. 2008). In addition, eastward phase propagation of precipitation and cloudiness is found downstream of the eastern TP (Asai et al. 1998; Wang et al. 2004, 2005; Yu et al. 2007b; Zhou et al. 2008; Chen et al. 2009). This diurnal phase propagation is similar to that observed leeward of the Rocky Mountains (Wallace 1975; Carbone et al. 2002; Carbone and Tuttle 2008). It is further found that phase propagation of diurnal cycles weakens along with the progress of the East Asian monsoon and almost ceases in midsummer (Asai et al. 1998; Wang et al. 2005; Chen et al. 2009). Wavelike propagation of afternoon convection, mountain–plain circulation, and nocturnal low-level jets are thought to be possible mechanisms responsible for the propagation and nocturnal rainfall (He and Zhang 2010; Huang et al. 2010).
Most of the above mentioned papers are focused on diurnal variations of precipitation and cloudiness by using just rain gauge or passive remote sensing observations (e.g., Asai et al. 1998; Wang et al. 2005; Yu et al. 2007b; Chen et al. 2009). Three-dimensional information of storms is never provided. It is hard to recognize what kind of storms (vertical structures) primarily contribute to the nocturnal rainfall in those studies. However, diurnal cycles of convection, properties of storms, lightning, and precipitation vertical structures are necessary for a better understanding of the underlying physical mechanisms, for evaluating rainfall retrievals, and for physical comparisons with model simulations (Nesbitt and Zipser 2003; Hirose and Nakamura 2005; Nesbitt et al. 2008; Liu and Zipser 2008).
The Tropical Rainfall Measuring Mission (TRMM) Precipitation Feature (PF) database combines comprehensive information from the Precipitation Radar, Microwave Imager, and Lightning Sensor together on the storm scale (Nesbitt et al. 2000; Liu et al. 2008a). This database is collocated well enough to investigate the phase difference between diurnal cycles of precipitation, deep convection, shallow convection, mesoscale convective systems (MCSs), and lightning (Nesbitt and Zipser 2003; Liu and Zipser 2008). It is now possible to measure how the vertical structures of precipitating systems vary diurnally by this database (Liu and Zipser 2008; Liu et al. 2008b). Furthermore, this feature-based database can provide unique insights into the phase propagation of the diurnal cycle from the perspective of precipitating storms.
Starting from that point, this study seeks to answer the following motivating questions: 1) Is there any phase propagation of diurnal variations of precipitating storms downstream from the eastern TP and how does it vary seasonally? 2) What are the phase differences among diurnal cycles of precipitation, deep convection, and lightning, and how do these change seasonally? 3) How does the precipitation vertical structure vary diurnally? 4) What are the possible mechanisms responsible for the phase propagation and nocturnal rain or convection?
During the warm season, the most evident change in the large-scale circulation and rainfall patterns happens with the onset of the mei-yu pattern (Tao and Chen 1987; Ding 1992). After the onset of mei-yu, deep southwesterlies from the tropics prevail over southern China and rainbands associated with the mei-yu front occur frequently (Chen 1994; Ding and Chan 2005; Chen 2004). The convection intensity and storm properties of mei-yu systems also vary from those of before onset (Xu et al. 2009). Therefore, this study focuses on the region of the eastern TP and continent downstream in different seasons: before the onset of mei-yu, the mei-yu season, and midsummer.
In this paper, details of the data and methods are described in section 2. Section 3 shows the 11-yr climatology of the large-scale flow, rainfall, and storm activity. Results of phase propagation of diurnal cycles, phase differences among different parameters, and their seasonal variations are presented in section 4. Comparisons between the afternoon and midnight peaks of rainfall and storm activity are given in section 5. Section 6 presents a summary showing how the changes in zonal wind profiles may be related to the phase propagation, and how the diurnal wind changes in the low levels may affect the diurnal cycles of rainfall. A brief summary of our principal conclusions is given in section 7.
2. Data and methods
a. ECMWF reanalysis data
The European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis Interim reanalysis data (ERA-Interim; Berrisford et al. 2009; Dee and Uppala 2009) between 1998 and 2008 are used to examine seasonal transitions and diurnal variations. The ERA-Interim dataset has a horizontal spatial resolution of 1.5° × 1.5° and 6-hourly temporal resolution. This dataset has 37 vertical levels ranging from 1000 to 1 hPa. Most of the levels are at 25-hPa intervals.
b. TRMM PF and 3B42 datasets
This study uses 11 yr (1998–2008) of the TRMM PF database (Nesbitt et al. 2000; Liu et al. 2008a). PFs are identified as contiguous near-surface raining areas derived from the TRMM Precipitation Radar (PR). Measurements from different instruments are collocated before grouping them into PR pixels with the adjacent-pixel method (Liu et al. 2008a). Standard products of 1B11, 1B01, 2A23, Lightning Imaging Sensor (LIS) orbital data, 2A25 (Iguchi et al. 2000), and 2A12 (Kummerow et al. 1998) are put into the collocation process. Parameters such as radar reflectivity, lightning, and microwave or IR brightness temperature can be derived directly from each pixel of the PFs at the resolution of the PR, that is, 4.2 km before and 5.1 km after the boost of the satellite in 2001.
The TRMM Multisatellite Precipitation Analysis (TMPA) 3B42 rain product (Huffman et al. 2007) is used to study the spatial distribution of seasonal rainfall and diurnal–nocturnal rainfall. The 3B42 product has 3-hourly temporal resolution and 0.25° × 0.25° spatial resolution, covering the globe from 50°S to 50°N, and is available from 1998 to 2008. This algorithm uses both passive-microwave measurements from low-earth orbit satellites and infrared radiance measurements from geostationary satellites. Multiple passive microwave rain estimates are first calibrated to TRMM PR and TRMM Microwave Imager (TMI) estimates before their combination. IR estimates are generated using the calibrated microwave estimates and are used to fill the passive microwave coverage gaps. The 3B42 rain estimates converge to TMI and PR estimates when they are available. Studies showed that 3B42 has good correspondence with rain gauge data for the rainfall amount and rainfall spatial pattern in China (Zhou et al. 2008; Shen et al. 2010).
c. Definition of total rain, MCS, and deep convection
Most of the studies on diurnal cycles focus on precipitation, but it is quite important to know the storm type, convection, and three-dimensional structure of precipitation systems contributing to the diurnal peaks. This study provides diurnal information on the structures of storms, and lightning, in addition to precipitation. Five parameters are selected and defined as follows (Table 1).
1) Population of PF
2) Population of MCS
An MCS is defined as a PF with area >2000 km2 having at least one convective pixel (Awaka et al. 1997). This definition guarantees an MCS having a mesoscale horizontal precipitating area with at least one convective cell embedded, as in the definition of Houze (1993). For this dataset, the population of MCSs is about 10% of the total PFs (Table 2).
3) Total rain
Total rain is defined by the total volumetric near-surface rainfall of a PF retrieved from the PR 2A25 algorithm (Iguchi et al. 2000). The total rain volume of a PF is the product of the area and mean rain rate of the feature.
4) Deep convection
Deep convection is defined by the area of the 20-dBZ radar echo at 12 km MSL (similar to Liu and Zipser 2008). By this definition, storms may or may not overshoot the tropopause (about 14 ∼ 15 km) but have precipitation-size ice particles lofted into the upper troposphere.
5) Flash counts
The total flash counts detected by LIS include both intracloud and cloud-to-ground lightning flashes within all the selected PFs (Nesbitt et al. 2000; Cecil et al. 2005). Although the number of total flashes may be caused by a single cell of a small thunderstorm or multiple cells of a large MCS, the presence of lightning is a well-known indicator of fairly intense convection (Cecil et al. 2005; Zipser et al. 2006).
d. Analysis methods
Generally, the East Asian summer monsoon has its early phase (mei-yu) first onset over South China in the middle of May (Chen 1983; Ding 1992). After the onset of mei-yu, the large-scale circulation and precipitation pattern change significantly from before (Ding and Chan 2005; Chen 1994; Xu et al. 2009). However, the westerly steering winds (300–500 hPa) at 20°–30°N are diminishing with the progress of the monsoon and vanish by mid-July (Murakami 1958; Murakami and Ding 1982; Chen 1993; Wang et al. 2005). As one of the major purposes of the study is to investigate the seasonal variability of the diurnal cycles and the phase propagation phenomenon, it is essential to consider the seasonal change of the tropospheric flow. Therefore, it is reasonable to define the first time period as “pre-mei-yu” during 1 April–11 May before the onset of the monsoon, the second as “mei-yu” during 15 May–25 June in the monsoon phase with significant steering winds, and the third as “midsummer” during 1 July–10 August when steering winds disappear.
The spatial distribution of the PF population is calculated in each 2° × 2° box, with the latitude-dependent sampling bias removed. The sampling bias is considered by calculating the difference of the total PR pixels over each box between 1998 and 2008. The bias is corrected by the bias factor defined as the fraction of the PR pixel number over each box to the mean PR pixel samples over the whole study region.
The region of eastern TP and the downstream region (23°–36°N, 90°–120°E) is first divided into three strips (dashed boxes in Fig. 1): strip 1 (32°–36°N), strip 2 (28°–32°N), and strip 3 (23°–28°N). Each strip is further separated into three boxes with different elevations (solid boxes in Fig. 1): 93°–103°E, 103°–112°E, and 112°–120°E. To describe the seasonal variations of the diurnal cycle, PFs are grouped into different seasons, strips, and regions. The PF samples in each strip and box during different seasons are listed in Table 2. The hour-dependent sampling bias of the TRMM PR for each box is also calculated and removed in a way similar to that used to correct for latitude-dependent sampling bias.
Spatial distributions of the diurnal cycles of PFs (or rainfall) over each strip are presented in Hovmöller diagrams similar to those of Carbone et al. (2002), but here they are done with 2-hourly and 2°-longitude bins. Diurnal cycles of occurrence frequency are derived from the summation of each defined parameter in eight local time bins (3-h bin) over each selected box during each specific season for 11 yr. In both the Hovmöller diagram and the frequency distribution diagram, the 1:2:1 filter is applied. The 1:2:1 filter is defined as the mean value of values in three consecutive hour–longitude bins with different weight [i.e., Ni = (Ni−1 + 2Ni + Ni+1)/4]. The diurnal cycle of the vertical precipitation structures are presented by the time–height contoured frequency by altitude diagrams (CFADs; 3-hourly and 1-km bin) of the occurrence frequency of the total area of the PFs with radar reflectivity ≥20 dBZ in each box. The occurrence frequency is unconditional. It is defined as the fraction of the total area to the total PR sampling area in a specific box.
In the study of nocturnal (2330–0530 LT) and afternoon (1130–1730 LT) rainfall and convection, the local time (LT) is defined as UTC plus 7 h. In the TRMM 3B42 dataset, the 3-hourly data point is defined as the total of the 1.5 h behind and ahead; for example, 0000 UTC is 2230–0130 UTC. Therefore, 2330–0530 LT covers both 1800 and 2100 UTC, while 1130–1730 LT covers both 0600 and 0900 UTC.
Using the ERA-Interim dataset, composite wind fields at 850 and 500 hPa during different seasons are generated separately to show the seasonal transition on the large-scale flow. The same ERA-Interim dataset is also used to construct vertical profiles of seasonal mean horizontal wind over the region (28°–32°N, 105°–115°E), showing diurnal phase propagation. Vertical cross sections of the U component at 0100 LT (1800 UTC) and 1300 LT (0600 UTC) are further created along the latitude of 30°N. Finally, the anomaly wind fields at 0100 and 1300 LT at 850 hPa are defined by the deviation of the 0100 and 1300 LT wind fields from the daily mean wind field.
3. Climatology of large-scale flow and rainfall
a. Large-scale flow at 850 and 500 hPa
Figure 2 presents the seasonal transitions of wind fields at 850 and 500 hPa over East Asia based on the ERA-Interim dataset. Seasonal transitions of the atmospheric flow are quite evident at both levels. Before the onset of the monsoon, anticyclonic circulation (the subtropical high) dominates over southeastern Asia, with westerly flow over the tropical South China Sea and southwesterly over southern China at both the 850- and 500-hPa levels. The subtropical ridge is located at about 20°N at 850 hPa and 15°N at 500 hPa and extends across the Indo-China Peninsula. At 500 hPa, strong westerlies prevail throughout East Asia north of 20°N.
After the onset of the monsoon, the subtropical anticyclonic circulation is replaced by strong south and southwesterly flow out of the deep tropics that brings a large amount of relatively warm and moist tropical air to the East Asian monsoon region. Also, the westerlies in the midtroposphere weaken and retreat northward. By midsummer, these westerlies retreat to northern China and the subtropical ridge progresses up to southern China. Although the southwesterly winds in the lower troposphere over southern China weaken during midsummer, they extend to northern China. One feature to note is that near the surface, easterlies flowing into the Sichuan basin, although persistent, weaken with the progress of the seasons. As can be seen in later sections and in the literature (Yu et al. 2009; Chen et al. 2010), the diurnal variations of the low-level wind change seasonally.
b. Rainfall and precipitation features
In general, the seasonal changes of the spatial distribution of rainfall from the 3B42 product and the precipitation features (Fig. 3) are consistent with that of the large-scale flow, especially at 850 hPa (Fig. 2). During pre-mei-yu, the heaviest rainfall is broadly distributed across southeast China between 22° and 30°N with a rainfall maximum of 300 mm month−1, and only 100 mm month−1 close to the foothills of the TP. During the mei-yu, rainfall over the South China Sea, the Bay of Bengal, and the foothills of TP increases markedly while remaining mostly south of the Yangtze River. The heaviest rainfall maxima located over southern China and Taiwan are mostly contributed by the mei-yu rainbands (Xu et al. 2009). Both rainfall and precipitation systems finally shift to the north of the Yangtze River with the northward penetration of the southwesterly flow in midsummer.
4. Diurnal variations of storm population, total rain, deep convection, and lightning
a. Downstream phase propagation1 and its geographic and seasonal variability
Generally, diurnal cycles of storm population and precipitation are dominated by afternoon maxima over high terrain, but in some seasons and locations they exhibit a phase delay, moving eastward into the lowlands, resulting in nocturnal maxima, except in midsummer (Figs. 4 –6). This confirms findings based on Geosynchronous Meteorological Satellite (GMS) retrievals by Asai et al. (1998) and Wang et al. (2005), and is similar to the phase propagation east of the Rocky Mountains during the warm season (Carbone et al. 2002). Another general feature is that the phase propagation is most evident in strip 2 in Figs. 4b and 5b, regions D and E during pre-mei-yu and mei-yu, but disappears in midsummer (Fig. 6b).
b. Phase differences of precipitation, deep convection, and lightning
The interpretation of the Hovmöller diagrams (Figs. 4 –6) showing phase propagation (or not) in certain locations and seasons can be clarified by also showing the diurnal cycles of the precipitation feature number, total rainfall, MCS number, flash counts, and deep convection, all of which are shown in Figs. 7 –9. There are too many interesting details to describe in this paper, so the following discussion is confined to the regimes that seem least ambiguous.
Before discussing the phase propagation (or its absence) in detail, we examine the relatively simple situation on the Tibetan Plateau (regions A and D). During all three seasons, these two regions have strong afternoon maxima in PF number, lightning flash counts, and deep convection, followed some hours later by maxima in MCS number and total rain (Figs. 7a,d, 8a,d, and 9a,d). This common sequence of events is strong evidence of the well-known local control of convection over the TP rather than large-scale dynamics (Uyeda et al. 2001; Qie et al. 2003). The time lag is reasonable for the organization of local convection into larger mesoscale systems, from which both convective and stratiform rain may fall for several hours.
The Sichuan basin shows a marked phase propagation during pre-mei-yu and mei-yu (Figs. 4e and 5e), all but disappearing in midsummer (Fig. 6e). However, the specific sequence of the peaking of each parameter is difficult to interpret in detail. The only clear result is that during pre-mei-yu and mei-yu, consistent with the notion of phase propagation of mesoscale phenomena eastward from the TP, all measures have nocturnal maxima, but they are not as sharp as those seen during the afternoon on the TP. The contrast in midsummer is striking (Fig. 9e). While there is no evident phase propagation in the total number of PFs (Fig. 6e), which actually have a weak afternoon maximum that suggests some locally generated convection, there are nocturnal maxima of deep convection, lightning, and early morning maxima of MCSs and total rain, suggesting that both deep convection and mesoscale systems have a mechanism favoring the nocturnal generation in the Sichuan basin, which is further explored in sections 5 and 6.
c. Diurnal variations of the vertical structure of precipitation
A great advantage of the TRMM Precipitation Radar data, in spite of the large attenuation at its 2-cm wavelength, is its ability to reveal vertical profiles of radar reflectivity, especially in the upper troposphere. If the attenuation correction (Iguchi et al. 2000) is applied through a large depth of strong echo, its accuracy does suffer, but the height reached by the 20-dBZ echo is not affected. Here, we construct the relative frequency of the radar echo area ≥20 dBZ at each altitude over selected regions and seasons (Figs. 10 and 11).
Over the Sichuan basin (box E), consistent with previous findings, extremely deep convection is present with 20 dBZ to 14 km from 0000 to 0600 LT during pre-mei-yu (Fig. 10a). But during mei-yu and midsummer, the data reveal a very different diurnal cycle in vertical structure. The first part of the early morning maxima still extends up to 15 km but this apparently evolved from an earlier (1800–2100 LT) peak. The second part of the early morning peak (0300–0600 LT) only goes up to 10 km during mei-yu and to 8 km during midsummer, perhaps indicating a predominance of decaying MCSs and stratiform rain, consistent with Fig. 9e.
Box G demonstrates a completely different situation. In all three seasons, there is a prominent late afternoon and evening peak of extremely deep radar echo, consistent with Figs. 7g, 8g, and 9g, and their lightning peaks, while much lower echo tops occur during early morning.
The diurnal cycles of the radar echo structure downstream of the above regions (boxes F and H) are different from E and G, but do not present any clear phase propagation from upstream. Multiple maxima are observed except during midsummer, when the afternoon peak of the very deep echo dominates. In all seasons, region F has an early morning peak with shallow echo tops, which, together with the morning rainfall maxima in Figs. 7f, 8f, and 9f, implies a role for decaying MCSs with stratiform rain. Region H has a complex structure in the pre-mei-yu, evolving toward a simple domination of an afternoon maximum in midsummer, also shown in Figs. 7h, 8h, and 9h.
5. Nocturnal versus afternoon peaks in rainfall, PF numbers, and lightning
a. Spatial distribution of rainfall
As mentioned above, the MCS population and total PR volumetric rain have nocturnal peaks over the foothills of the TP and over central eastern China in certain seasons. The spatial distribution of nocturnal rainfall based on TRMM 3B42 rainfall data (Fig. 12) is quite consistent with the above findings and with previous studies (Yu et al. 2007b; Li et al. 2008). During pre-mei-yu, there is a band-shaped nocturnal maximum extending from the Sichuan basin to the coast. In contrast, afternoon rainfall maxima exist along and within 500 km of the coast of south and central China, with a prominent minimum in the Sichuan basin. The east–west nocturnal rainfall center may be related to the propagation of precipitation systems downstream of the eastern TP as demonstrated previously.
The band-shaped nocturnal maximum downstream of the eastern TP is less evident during mei-yu and disappears in midsummer. Nocturnal rainfall maxima still appear over the south and southeast foothills of the TP, but closer to the foothills. The spatial pattern of afternoon rainfall is quite similar in mei-yu and midsummer, except that significant rainfall extends northward and spreads through all of eastern China, while the afternoon minimum persists in the Sichuan basin.
b. Distribution of PFs and lightning features
Spatial distributions of the occurrence of precipitating features during 2330–0530 LT versus 1130–1730 LT during different seasons are presented in Fig. 13. Figure 14 shows the distribution of PFs by numbers of lightning flashes during selected periods in different seasons. Over the TP, afternoon storms dominate in every season, similar to the classic diurnal characteristics over high terrain (Fig. 13).
In the foothills of the southeastern TP and eastward for 1000 km, higher percentages of nocturnal PFs occur, especially during the pre-mei-yu. Many of these nocturnal precipitation features have lightning of moderate to high flash rate, indicating either intense convection, large mesoscale convective systems, or both. But during mei-yu this nocturnal peak is less obvious. Lightning downstream of the TP is even less frequent in the afternoon during mei-yu than nocturnal lightning during the pre-mei-yu. This seasonal difference in lightning activity has been described in Xu et al. (2010) and is consistent with their finding of higher radar reflectivity from the TRMM PR in the mixed phase region.
6. Discussion: Possible mechanisms for the variability of phase propagation and diurnal cycles
Previous sections show that the nocturnal and afternoon peaks of rainfall and other parameters show both different seasonal and geographic distributions. Here, we ask how the seasonal changes in zonal wind profile and how the diurnal cycle in the low-level flow field may help explain the more striking features of the rainfall cycle, most notably, the dominance of the nocturnal and early morning rainfall in the Sichuan basin.
a. Relating the phase propagation to the zonal wind profile
The most obvious phase propagation has been shown to be in strip 2, the latitude belt from 28° to 32°N extending from the high TP through the Sichuan basin and eastward. To supplement the large-scale flow fields at just two levels (Fig. 2) for this region, Fig. 15 shows the mean zonal wind profile for the region of strongest phase propagation for each season. Both pre-mei-yu and mei-yu have substantial zonal wind components above the 700-hPa level, as well as strong wind shear, especially in pre-mei-yu. In midsummer, the mean zonal wind in this belt essentially disappears.
Carbone et al. (2002) and others have demonstrated a recurrent pattern of phase propagation of rainfall maxima in the central United States downstream of the Rocky Mountains, related to the maintenance of deep convection and mesoscale systems, initially phase locked to afternoon convection over the Rockies, resulting in nocturnal maxima well downstream, moving in the general direction of “steering-level winds” near 700 hPa, but typically about 8 m s−1 faster. It is well beyond the scope of this paper to distinguish between the many possible reasons for a specific speed with respect to the wind at any particular level. But we do note that the wind profile in and east of the Sichuan basin (Fig. 15) is fairly close to that of Carbone et al. (2002). Further, the observed phase propagation in the Sichuan basin is between about 12 and 15 m s−1 in mei-yu (from Figs. 4e and 5e) compared with Carbone et al.’s 14 m s−1. It is an open question whether our results should be interpreted as having a phase speed 10–12 m s−1 faster than our 700-hPa wind, or whether our steering-level winds should be taken closer to 600 hPa (the TP is higher than the Rocky Mountains), in which case our results also imply a phase speed about 8 m s−1 faster.
b. Relating the diurnal cycle of precipitation features and rainfall to the diurnal cycle of low-level winds
Despite the rather coarse 1.5° resolution of the ERA-Interim reanalysis data used in this paper, the changes between the 1800 UTC (0100 LT) and 0600 UTC (1300 LT) low-level wind fields are sufficiently large to help explain the diurnal cycle of the observed rainfall and precipitation systems. Figure 16 shows the west–east vertical cross section of the zonal winds through the center of the Sichuan basin for each season, while Fig. 17 shows the anomaly wind fields from the seasonal mean for 0100 and 1300 LT at 850 hPa for a region that includes the Sichuan basin.
Above 700 hPa, there is little change in the tropospheric zonal wind profile between 0100 and 1300 LT, but there is a large change in the low-level winds. Specifically, the low-level easterly flow near 850 hPa converges in the center of the Sichuan basin at 0100 LT in all seasons, near 105°E, but at 1300 LT in all seasons that convergence is along the slope of the Tibetan Plateau near 103°E. This convergence of about 1 m s−1 may seem small, but it is consistent with rising motion on the slopes during the afternoon, and rising motion in the center of the basin during the night. Using the slightly better resolution of 1.0° of the National Oceanic and Atmospheric Administration (NOAA) Global Forecast System (GFS), He and Zhang (2010) demonstrated a very similar magnitude of diurnal wind change, consistent with a similar phase propagation and diurnal cycle of rainfall along the terrain slope in northern China.
Several authors [including Carbone et al. (2002) and He and Zhang (2010)] have pointed to the existence of a nocturnal low-level inflow of moist air that may assist the formation and maintenance of the nocturnal precipitation in many regions of the world. The anomaly wind field in this case (Fig. 17) supports this interpretation in the Sichuan basin as well. In all three seasons, the 850-hPa flow changes from southerly to northerly with a vector change of about 2 m s−1 between 0100 and 1300 LT. In addition, the change from convergence to divergence between 0100 and 1300 LT is obvious, and at least as large a magnitude as that shown by He and Zhang (2010) for their study region in northern China (their Fig. 5). While the absolute wind speeds hardly qualify for the label “low-level jet” in the seasonal mean, the low-level flow change is consistent with a more favorable flow of potentially unstable air into the Sichuan basin at night.
The principal findings of this study include the following points:
There is a phase propagation of diurnal cycles of precipitating storms, total rainfall, convection, and lightning from the foothills of the eastern TP downstream, which is most evident over the Sichuan basin during the pre-mei-yu and mei-yu seasons, absent during midsummer, and consistent with a lack of zonal winds and zonal shear during midsummer.
Before midsummer, the eastern TP foothills are dominated by nocturnal rainfall, but the early morning peak of precipitation is only in phase with deep convection, MCSs, and lightning during pre-mei-yu.
Most of the nocturnal precipitation is in phase with MCSs and possibly contributed to by long-lived MCSs evolving from late afternoon or early night convection, but these early morning MCSs show larger percentages of deep convection in pre-mei-yu.
In midsummer, most of East Asia is dominated by afternoon precipitation and convection, with the Sichuan basin and northern part of eastern China as the exceptions where early morning rainfall prevails, in spite of most deep convection and lightning occurring during the afternoon. Most of the midnight and early morning rainfall is contributed by MCSs.
In all three seasons, there is strong low-level convergence in the Sichuan basin during night time with southerly winds flowing into the basin, consistent with the nocturnal maximum of rainfall, while divergence during the day is consistent with a marked rainfall minimum.
This research was supported by NASA Precipitation Measurement Mission Grant NAG5-13628 under the direction of Dr. Ramesh Kakar. Special thanks are given to the TRMM Science Data and Information System (TDSIS) group led by Drs. Erich Stocker and John Kwiatkowski at NASA Goddard Space Flight Center, Greenbelt, Maryland, for data processing assistance. Many thanks also go to Dr. Chuntao Liu at the University of Utah, Salt Lake City, Utah, for science discussions. Careful and constructive reviews by Dr. Richard Carbone and two anonymous reviewers resulted in significant improvements to the manuscript and are greatly appreciated.
Corresponding author address: Weixin Xu, Dept. of Atmospheric Sciences, University of Utah, 135 S 1460 E, Rm. 819, Salt Lake City, UT 84112-0110. Email: email@example.com
In this paper, phase propagation refers to the apparent total phase speed in the zonal direction, without implying anything about storm propagation with respect to the wind at any level.