1. Introduction
During the transition between the cold and warm seasons from May to July, a rainy mei-yu (plum rain, baiu, or jangma) season takes place in East Asia (e.g., Ray et al. 1991; Lin et al. 1992; Ding 1992; Chen and Liang 1992). In Taiwan, the mei-yu season usually lasts for a month from mid-May to mid-June. The main weather system during this time is the mei-yu front, which is a quasi-stationary front associated with a significant water vapor gradient and strong southwesterly (hereafter SW) winds on the south side (Lin et al. 1992; Li et al. 1997). Most of the mei-yu fronts observed during the Taiwan Area Mesoscale Experiment (TAMEX; Kuo and Chen 1990) and the Terrain-influenced Monsoon Rainfall Experiment (TiMREX; Tu et al. 2014; Wang et al. 2014) exhibit baroclinic characteristics over southeastern China (Chen et al. 1989; Trier et al. 1990; Chen and Hui 1990, 1992), but the frontal structure becomes shallower after advancing to the vicinity of Taiwan (Chen et al. 1989; Tu et al. 2022). Some frontal systems can develop after the eastward/southeastward movement of a lee vortex that forms on the leeside of the Tibetan Plateau (Chen and Chen 1995), which is commonly called the southwest vortex (SWV)(e.g., Feng et al. 2016; Kuo et al. 1988). The SWV is usually associated with strong SW winds [so-called low-level jet (LLJ)] in the low levels (850 or 700 hPa) in its southeastern flank (Chen and Yu 1988; Chou et al. 1990; Chen and Chen 1995; Hsu and Sun 1994). The SW winds can also cover a rather large area, and when they do, the case can be regarded as a SW flow event (Chien and Chiu 2019; Chien et al. 2021). During either a LLJ or SW event, warm moist air is transported by the strong SW winds from the tropical ocean to the frontal area and is lifted by the front, leading to the formation of heavy rainfall near the front (Chen et al. 2008, 2005; Chen and Yu 1988; Chien 2015).
The amount of precipitation during a mei-yu season has great interannual variability in East Asia, which may be affected by several factors, including the interaction of water vapor transport between the tropics and midlatitudes, the strength of the extratropical jet stream, and the strength of the western North Pacific subtropical high (hereafter PSH; Volonté et al. 2022). A positive feedback during the interactions among the SW flow, water vapor transport, and mei-yu fronts may also affect the amount of precipitation in a mei-yu season in Taiwan (Chien et al. 2021). In addition, changes in sea surface temperature during El Niño–Southern Oscillation (ENSO) may affect the timing of the monsoon over the South China Sea (SCS), and consequently the associated precipitation, through equatorial Rossby waves (Lu et al. 2020).
The mei-yu season of 2020 was one of the most extreme rainy seasons in East Asia. The accumulated rainfall in the Yellow River basin of China from June to July 2020 broke the record held since 1961 (Takaya et al. 2020; Guo et al. 2021). This was primarily affected by the interdecadal increase in the sea surface temperature of the tropical Indian Ocean. The PSH was thus enhanced, resulting in a stronger southwest monsoon and moisture flux convergence in the vicinity of the Yellow River. The accumulated rainfall in the Kyushu region of Japan in early July of 2020 also broke the previous 30-yr record. Several factors were found to be related to the extreme event, including a nearly stationary high-level trough, a stronger and more southward upper-level jet, and the Silk Road wave that started in late June (Horinouchi et al. 2021; Hirockawa et al. 2020; Araki et al. 2021). Under the influence of such combined effects, the convergence of moisture flux lasted for dozens of days, resulting in continuous rainfall near Kyushu.
In Taiwan, the first half of the mei-yu season (15–31 May, hereafter FHMY) in 2020 was also a record-breaking period of extreme precipitation. In this period, the average accumulated precipitation of the 28 surface stations of the Central Weather Bureau (CWB) in Taiwan was the largest in 42 years, from 1979 to 2020 (Fig. 1a). The total amount of 345.3 mm was about twice the climatological mean (168.3 mm) and exceeded a positive anomaly of 2.1 standard deviations. During the six 6-h periods from 0600 UTC 21 May to 1200 UTC 22 May, in particular, all and four of them had a rain intensity exceeding the 95th and 99th percentiles of the climatology, respectively (Fig. 1b). In the FHMY climatology of 42 years, there were only two cases that had rain intensity exceeding the 95th percentile continuously for at least 36 h. The current case was the longest continuous precipitation case that exceeded the 99th percentile of the climatological precipitation intensity four times out of six. The other case that occurred around 21 May 2014 exceeded the 99th percentile only one time. During the 36 h, 135.9 mm of precipitation on average was accumulated in Taiwan, which undoubtedly set a record in FHMY. The maximum average rain intensity reached 30.1 mm (6 h)−1 at 0600 UTC 22 May, which was close to the climatological extreme of FHMY in Taiwan. These findings indicate that the case presented in this paper is a climatologically rare event of persistent extreme precipitation.
(a) Accumulated average rainfall (mm) of the 28 surface stations of the CWB (see Fig. 4a for locations) during FHMY (15–31 May) in 2020 (red) and 1979–2019 (green), with the climatological mean from 1979 to 2020 shown in blue. The total rainfall amounts during FHMY (with standard deviation) are denoted on the top left of the diagram. (b) Average 6-h rainfall intensity in Taiwan [unit: mm (6 h)−1] for the same period in 2020 (red), with a blue line showing the maximum and three black lines from top to bottom representing the 99th, 95th, and 90th percentiles of the climatological rain intensity, respectively. (c) Red dots denote the occurrence of the SW flow event in southern Taiwan [SWs; see Chien et al. (2021) for a definition] during FHMY in 2020.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
The satellite imaginary at 0000 UTC 22 May 2020 shows a long and wide cloud band extending westward from the ocean southwest of Taiwan to the Indo-China Peninsula (Fig. 2a). Many strong mesoscale convective systems (MCS) associated with high clouds developed vigorously in the cloud band. The 850-hPa geopotential height (Fig. 2b) from the ERA5 (the fifth generation of atmospheric reanalysis, Hersbach 2016) data of the European Centre for Medium-Range Weather Forecasts (ECMWF) shows that at this time, a low pressure system extended from South China to the north side of Taiwan. This low increased the pressure gradient to its southeast side, resulting in strong SW winds over a large area from the northern SCS to the ocean east of Taiwan. The low is associated with the SWV that originated over the southeastern flank of the Tibetan Plateau in southwest China 2–3 days earlier and moved east-southeastward (Fig. 2c). At the surface (Fig. 2d), a mei-yu front associated with the pressure trough and strong wind convergence extended from southern China, through the central Taiwan Strait, and northeastward to the east of northern Taiwan. The MCSs were triggered over the upstream ocean when the SW flow was lifted by the mei-yu front. They moved into the Taiwan area (e.g., Lau et al. 1988; Kuo and Chen 1990; Zhang et al. 2003) and were enhanced by the Central Mountain Range (e.g., Akaeda et al. 1995; Teng et al. 2000; Chen et al. 2006; Xu et al. 2012; Tai et al. 2020), resulting in the extreme precipitation event in Taiwan. The 12-h accumulated rainfall (Fig. 3a) shows that there was not much rain in Taiwan before 0000 UTC 21 May. Precipitation started to increase in the daytime (0000–1200 UTC) of 21 May as the SWV approached northern Taiwan. The amount of rainfall kept increasing, primarily in southern Taiwan, from 1200 UTC 21 May to 1200 UTC 22 May, and then decreased during the last period. Affected by the SW flow and the front, heavy precipitation mainly occurred in the low-plain and mountainous areas of southwestern Taiwan. To understand the meteorological factors that led to this heavy rain event, we compared the environmental conditions of FHMY in 2020 with climatology. In addition, an ensemble experiment of model simulations was conducted to examine the contributions of these factors to the rainfall in a more quantitative way. The key questions to be addressed in this paper are as follows:
-
Why is the rain intensity strong and the period of extreme precipitation long in FHMY of 2020?
-
How do the environmental fields of FHMY in 2020 differ from climatology?
-
What are the key factors that led to heavy precipitation from 21 to 23 May 2020?
Fig. 2. (a) Himawari infrared cloud imagery at 0000 UTC 22 May 2020. (b) 850-hPa wind vector (m s−1), wind speed (color shading; m s−1), and geopotential height (contours; interval: 10 gpm) from the ERA5 data at 0000 UTC 22 May 2020. (c) Hovmöller diagram of the 850-hPa geopotential height anomaly (color shading; gpm) and relative vorticity (green contours; interval: 3 × 10−5 s−1) along line AA′ [location shown in (b)]. Thin solid (dashed) contours represent positive (negative) values, while thick contours denote the zero line. The ordinate is time from 1200 UTC 19 May to 0000 UTC 23 May 2020. (d) As in (b), but for 10-m winds and sea level pressure (contours; interval: 2 hPa), with the mei-yu front noted.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
Fig. 3. (a) 12-h accumulated rainfall (mm) from rain gauge stations (black dots) in Taiwan ending at (from left to right) 0000 UTC 21 May, 1200 UTC 21 May, 0000 UTC 22 May, 1200 UTC 22 May, and 0000 UTC 23 May 2020. The number in the lower-right corner denotes the maximum rainfall in each panel. (b) As in (a), but for the ensemble mean of the 64 members. (c) Time series of the average rainfall intensity (mm h−1) over land area of box A [shown in (a)] from the 64 ensemble members. The solid line, shaded area, and dots denote the mean, the one standard deviation range, and the extreme value of the ensemble, respectively.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
The data sources, experimental design, and model settings are presented in section 2. In the third section, composite analyses of the reanalysis data are used to identify the difference between the environmental fields of 2020 and the climatology. Section 4 presents the ensemble simulation results to clarify the factors that led to the heavy rainfall. Last, a summary is presented in section 5.
2. Data and experiment design
This paper uses the ERA5 data provided by the ECMWF for the analyses of climate statistics during FHMY from 1979 to 2020. The horizontal spatial resolution of the data is 0.5° × 0.5°, and the time resolution is every 6 h. If the atmospheric state at every 6 h is labeled as an event (or sample), the total amount of samples evaluated in this study is 2856.
The methods of defining the SW flow event and calculating the observed precipitation intensity in Taiwan are the same as those of Chien et al. (2021) and are briefly described below. Basically, the 850-hPa winds from the ERA5 are averaged in fourteen 2° × 2° boxes surrounding Taiwan (Fig. 4a). A SW flow event is identified if it meets the following two criteria: 1) at least one box is required to have a mean wind speed greater than 12 m s−1 with a SW wind direction, and 2) at least half of the 14 boxes are required to have a SW wind direction. The hourly rain data from the 28 surface stations of the CWB (Fig. 4a) are utilized to calculate the average rainfall intensity [unit: mm (6 h)−1] in Taiwan.
(a) The 14 boxes (2° × 2°) surrounding Taiwan are the areas for wind average, with numbers showing centers of the boxes. Color shading is terrain height, and 28 blue dots are locations of the CWB surface stations. (b) Domain setting of the model, which includes three nested domains with 27-, 9-, and 3-km horizontal resolutions. (c) Time frames of the ensemble simulations. The ETKF DA is performed nine times (eight cycles) in domain 1 from 1200 UTC 18 May to 1200 UTC 20 May 2020. The 64 ensemble member runs in the three nested domains are initialized at 1200 UTC 20 May and end at 0000 UTC 23 May 2020.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
Using the ERA5 data, this paper computes averages of meteorological fields for four composite groups. The first group is CM (climatological mean), which takes a climatological mean of FHMY from 1979 to 2020 using all the aforementioned 2856 samples. The second group is Y20, which is the average of 68 samples from 15 to 31 May 2020. Representing the climatological mean of heavy precipitation events, the R90 group takes the average of the 286 samples that have a rainfall intensity greater than, or equal to, the 90th percentile of the climatological rainfall intensity [7.6 mm (6 h)−1]. Last, the Y20R group is the average of 6 samples from 0600 UTC 21 May to 1200 UTC 22 May 2020, which set a record for the longest continuous heavy rain events in Taiwan during FHMY in 42 years (Fig. 1b).
The Weather Research and Forecasting (WRF) Model version 4.1.2 (Skamarock et al. 2019) is used for the ensemble simulations. The domain setting includes 3 nested domains with 27, 9, and 3 km horizontal resolutions (Fig. 4b), and 45 levels in the vertical. A two-way interaction is applied between two adjacent nested domains. The Goddard scheme (Tao et al. 1989) and the Yonsei University scheme (Hong et al. 2006) are utilized for the microphysics and the boundary layer processes, respectively. Two cumulus parameterization schemes (CPS): Tiedtke (Tiedtke 1989) and Betts–Miller–Janjić (Janjić 1994) schemes, are used during data assimilation (DA) process, while only the Betts–Miller–Janjić CPS is used during the model simulation. The CPS is only applied in domains 1 and 2. The ERA5 data with a horizontal spatial resolution of 0.25° × 0.25° and a time resolution of 6 h are used for the initial and boundary conditions of the model.
Applying the method developed by Bishop et al. (2001) and using the inflation factor strategy described by Wang et al. (2007), the ensemble transform Kalman filter (ETKF; Barker et al. 2012) DA with a 32-member ensemble is performed nine times (or eight cycles) at a 6-h interval from 1200 UTC 18 May to 1200 UTC 20 May 2020 (Fig. 4c), during which only domain 1 is run. The above DA process is repeated twice using the aforementioned two different CPS schemes to create a 64-member ensemble. The assimilated data include traditional observations from surface stations, soundings, aircraft, and ships. During the 60-h forecasting period from 1200 UTC 20 May to 0000 UTC 23 May 2020, 64 ensemble member runs are carried out in the three nested domains.
3. The environmental conditions of FHMY in 2020
The warm-season precipitation in Taiwan is significantly affected by the SW flow (Chien and Chiu 2019; Chien et al. 2021), and there is a high correlation between the two in terms of both interannual variations and short-term temporal changes. In FHMY of 2020, the SW flow events in southern Taiwan identified by the method of Chien et al. (2021) occurred 15 times (Fig. 1c), which was about 3.8 times the climatological mean (4) and was the second most in 42 years. Comparisons between Figs. 1b and 1c also show a high correlation between the SW flow occurrence and heavy rain. It is thus clear that during FHMY of 2020 the SW flow, in addition to mei-yu fronts, played an important role in heavy rainfall near the Taiwan area.
In the CM group, the PSH gradually intensifies from spring to the early mei-yu season, resulting in an increase in the northwest–southeast geopotential height gradient near Taiwan in FHMY (Fig. 5a). During this time, the East Asian summer southwest monsoon is starting to be established (Wang 2006; Ding and Chan 2005). The main IVT (vertically integrated water vapor transport from 1000 to 200 hPa, Ralph et al. 2019) moisture conveyor belt (Fig. 5b) extends from the Indian Ocean, through the southern side of the Indo-China Peninsula, and to the northern SCS. This belt, when it is stronger, can provide a source of moisture and creates a warm, humid environment in the vicinity of Taiwan, causing the equivalent potential temperature gradient to increase in the early mei-yu season. Mei-yu fronts associated with MCSs form more frequently to the north and move closer to Taiwan under these favorable conditions (Thomas and Schultz 2019; Chien et al. 2021). In the middle atmosphere (Fig. 5c), upward vertical motion associated with the low pressure environment of the mei-yu front extends from South China to the south of Japan.
(a) 850-hPa geopotential height (contours; interval: 10 gpm), wind (vectors; m s−1), and specific humidity (color; g kg−1); (b) 850-hPa equivalent potential temperature (color; K) and the IVT (vectors; kg m−1 s−1); and (c) 500-hPa geopotential height (contours; interval: 20 gpm), wind (vectors; m s−1), and vertical velocity (color; unit: 0.01 Pa s−1) in the climatological mean (CM). The scale of color shading of each row is denoted in the rightmost column. Gray indicates the area below topography. (d)–(f) As in (a)–(c), but for Y20.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
The PSH in Y20 was significantly stronger and extended more southwestward near the southern Indo-China Peninsula (Fig. 5d) than in CM (Fig. 5a), considering the 1510-gpm isoheight as an example. As a result, the geopotential height gradient increased and the SW wind increased significantly over the SCS. Moreover, a super cyclone named Amphan moved north-northeastward over the Bay of Bengal on 17–20 May and made landfall on West Bengal at 1200 UTC 20 May 2020. This storm caused the airflow to turn northeastward at the western Indo-China Peninsula in Y20 (Figs. 5d–f). Compared with that in CM, the moisture conveyor belt (large IVT) in Y20 was stronger over the northern Indo-China Peninsula with higher equivalent potential temperature (Fig. 5e). Since both the wind speed and water vapor in Y20 were higher than those in CM, the meteorological environment near Taiwan during FHMY of 2020 was significantly warmer and wetter than in the climatology. The equivalent potential temperature gradient increased significantly to the north of Taiwan (Fig. 5e), which was conducive to the increased frontal activity, SW flow generation, and MCS development near Taiwan (Fig. 5f). The FHMY of 2020 thus became the most extreme precipitation season during FHMY in 42 years.
The intensity of the PSH in R90 (Fig. 6a) is not much different to that in CM (Fig. 5a). However, the low pressure system in eastern China and the intensified midlatitude trough over Korea and Japan cause an increase of the geopotential height gradient and the SW wind over a large area extending from the SCS, through southern Taiwan, and northeastward to the south of Japan. In the same region, the IVT is large (Fig. 6b). High equivalent potential temperature covers the SCS and the Taiwan area. At 500 hPa, the midlatitude trough extends from Korea southward to the north of Taiwan (Fig. 6c). Compared with CM, R90 has larger upward motion associated with stronger convection over a more southward region extending from the SCS to the south of Japan. The vicinity of Taiwan is influenced by warm, humid weather conditions that favor the activity of mei-yu fronts and the formation of precipitation. These environmental conditions are under a typical configuration when the East Asian mei-yu front is lingering around Taiwan (Chen 1983, 1992). To summarize, the climatologically heavy precipitation events in FHMY are mainly caused by the mei-yu front and the SW flow on the south side of the front.
As in Fig. 5, but for (a)–(c) R90 and (d)–(f) Y20R.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
Compared with R90 (Fig. 6a), Y20R (Fig. 6d) shows that the PSH was more intense and the low pressure system over China shifted more southward and became stronger over southeastern China. As a result, the geopotential height gradient of Y20R in the vicinity of Taiwan was significantly larger than that of R90, leading to SW flow events with fairly strong winds near Taiwan during these 36 h. Compared with that of Y20 (Fig. 5d), the PSH of Y20R (Fig. 6d) was slightly weaker. However, the low pressure system over the Bay of Bengal due to Cyclone Amphan was stronger and shifted to a more northward location in Y20R than in Y20. As a result, strong SW airflow crossed the northern Indo-China Peninsula into southern China and turned westerly over the southern side of the Tibetan Plateau, evident from the airflow at 850 hPa (Fig. 6d) and 500 hPa (Fig. 6f) and the IVT (Fig. 6e). A SWV tended to form after this strong westerly flow moved to the low-plain region of southwestern China (Feng et al. 2016; Shu et al. 2022). Therefore, the low to the northwest of Taiwan in Y20R (Fig. 6d) was associated with the SWV discussed in Fig. 2. The strong west-southwesterly flow associated with the SWV could continuously transport moisture from South China and the Indo-China Peninsula (Figs. 6d,e) toward Taiwan, creating favorable conditions for the development of strong MCSs and heavy precipitation in regions of upward lifting (Fig. 6f).
To determine how extreme the SW flow and its associated moisture flux of Y20R were, we used the ERA5 data to compute the areal mean of the 850-hPa wind and moisture flux in a 2° × 2° box southwest to Taiwan (box 13 in Fig. 4a) for all samples in the CM group. Figure 7a shows that most of the Y20 samples had stronger SW flow than the climatological median. Six of them, which belong to the Y20R group, were close to or surmounted the top 1% (the 99th percentile) of the climatology, with areal-mean wind speeds all exceeding 15 m s−1. The time sequence of these six samples shows that the SW flow turned to a more west-southwesterly direction at the time of the third sample (1800 UTC 21 May), and then increased its intensity at 0000 and 0600 UTC 22 May, with the strongest wind reaching 20 m s−1. At the last time of the period, the SW flow weakened and changed to more westerly. Moisture flux in Fig. 7b shows a similar result, except that there was one more sample in the Y20R group exceeding the top 1% of the climatology. It is thus evident that Y20R is an extreme case of strong SW flow events, associated with not only intense low-level winds, but also plenty of moisture.
Areal-mean 850-hPa (a) wind and (b) moisture flux in box 13 of Fig. 4a (21°–23°N, 118°–120°E) for all the 2856 samples of the CM group (blue dots) from 1979 to 2020, using the ERA5 data. The 68 and 6 samples of the Y20 and Y20R group are marked as green and red dots, respectively. Numbers in red denote the time sequence of the 6 Y20R samples. The abscissa and ordinate is the x and y component, respectively, of the (a) wind and (b) moisture flux. Three circles from center to outside represent the median, 90th, and 99th percentiles of the climatology, which are 5.5, 10.7, and 16.9 m s−1 for wind, and 65.1, 138.4, and 235.0 m s−1g kg−1 for moisture flux, respectively. The number of outliers are shown in the lower-right corner.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
4. Favorable conditions for heavy rainfall
The mean rain of the 64 ensemble members (Fig. 3b) shows that rain started to increase at 0000–1200 UTC 21 May and reached its peak at 0000–1200 UTC 22 May 2020. Most precipitation occurred in the southern part of Taiwan, with the maxima located in the mountainous areas. From 1200 UTC 22 May to 0000 UTC 23 May, rain gradually decreased. Although rainfall in northern Taiwan was underforecasted, the ensemble overall captured the main precipitation evolution and the rain distribution in central and southern Taiwan during the 60-h forecasts, compared with the observation (Fig. 3a). Time series of the average precipitation in southern Taiwan (over land area in box A of the leftmost panel in Fig. 3a) from the 64 ensemble members (Fig. 3c) show two rainfall peaks around 0000 UTC 22 May, which are similar to those in the observation (Fig. 1b). To analyze how the environmental fields affect the precipitation intensity during the 36 h of Y20R, this paper selects two 6-h periods associated with the two rainfall peaks. The first period (hereafter P1) is from 2100 UTC 21 May to 0300 UTC 22 May, while the second period (hereafter P2) is from 0600 to 1200 UTC 22 May.
a. Environmental conditions during heavy rainfall
During P1, the ensemble-mean rain of the 64 members showed large rainfall over the mountain in southern Taiwan, with a maximum of 146 mm (6 h)−1, and smaller rainfall over central Taiwan (Fig. 8b). These two heavy rainfall areas and the overall rainfall pattern in Taiwan agreed well with the observation (Fig. 8a), except for less rainfall over the southwestern low plains in the simulation. Standard deviation of the ensemble members was in general small (Fig. 8c). The ensemble-mean rainfall pattern in P2 (Fig. 8e) was overall similar to that of P1, except that rainfall increased slightly in southern Taiwan and decreased in central Taiwan. Compared with the observation (Fig. 8d), rainfall in southern Taiwan was underforecasted, and that in northern Taiwan was not reproduced (Fig. 8e). The standard deviation of the ensemble members in P2 became larger than in P1, especially in southern Taiwan (Fig. 8c). The fact that the ensemble mean overall had more rainfall in the mountain areas than the low-plain regions resembles a common model bias in simulating rainfall over complex terrain. In case of rain forecasts from individual members, however, some members in the ensemble did produce better rainfall than others (as discussed later). Considering this fact and the challenge of the model in forecasting the correct strength, size, and timing of MCSs, the 6-h ensemble-mean rainfall shown in Figs. 8b and 8e was somewhat acceptable.
6-h accumulated rainfall (mm) from (a) the rain gauge stations and (b) the ensemble mean of the 64 members during P1 (2100 UTC 21 May–0300 UTC 22 May 2020), with (c) the standard deviation of the ensemble also shown. The number in the lower-right corner denotes the maximum value of the panel. (d)–(f) As in (a)–(c), but for P2 (0600–1200 UTC 22 May 2020). Box A in (a) is as in Fig. 3a.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
Figure 9 shows the probability distribution function (PDF) of the average 6-h accumulated rainfall over land area of box A in Fig. 8a from the 64 ensemble members. In P1, 37 ensemble members predicted rainfall smaller than 60 mm (6 h)−1, 11 members between 60 and 70 mm (6 h)−1, and 16 members larger than 70 mm (6 h)−1 (Fig. 9a). Since the average rainfall in observations was 66.3 mm (6 h)−1, the ensemble in general underforecasts the rainfall in P1. During P2, the simulated rainfall was more evidently underforecasted (Fig. 9b), with a lot more members having rainfall smaller than the observation of 77.1 mm (6 h)−1.
Probability density distribution (unit: number of members) of the average 6-h precipitation (mm) over the land area of southern Taiwan in box A of Fig. 8a from the simulations of the 64 ensemble members during (a) P1 and (b) P2. Black dashed lines represent observations.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
The 850-hPa geopotential height, mixing ratio, and winds from the ensemble mean of the 64 members show that when the SWV was still located over southeastern China and far from Taiwan at 0000 UTC 21 May (Fig. 10a), the large-scale monsoonal SW flow with large water vapor flux was extending from the northern SCS to southern Taiwan. A pressure trough associated with weak cyclonic winds was present over the southern Taiwan Strait at 1000 hPa (Fig. 11a). The 1000-hPa virtual potential temperature gradient (|∇θυ|) was weak over the Strait (Fig. 11b). Rainfall started to occur in southern Taiwan at this time (Fig. 3b). However, since there was not much low-level convergence in the upstream flow, the accumulated precipitation was small. At 2100 UTC 21 May, the starting time of P1, the eastern portion of the SWV had arrived at the north side of Taiwan (Fig. 10b). The strong west-southwesterly winds associated with the SWV and the SW winds of the large-scale monsoonal flow transported large moisture toward the Strait. At 1000 hPa, a low associated with the SWV appeared over the northern Strait (Fig. 11c). Many large convergence regions extended from the low center southwestward to the southern Strait alongside the streamline convergence of the ensemble-mean winds. A similar situation was also found in the |∇θυ| pattern (Fig. 11d). Since divergence and |∇θυ| are computed in each individual member before an ensemble mean is taken, these areas correspond to the frontal locations in individual ensemble members. The values of convergence and |∇θυ| within the concentrated frontal zone are actually a lot larger in individual members than the ensemble mean. At this time, fronts were in general stronger than at the early times (e.g., Fig. 11b), evident by the more compact contours of the mean θυ (Fig. 11d). When the moisture-laden SW flow was lifted by the mei-yu front, MCSs formed over the ocean and moved over land, resulting in heavy rainfall in southern Taiwan (Fig. 8b). At 0600 UTC 22 May, the starting time of P2, the center of the SWV had passed the northern tip of Taiwan (Fig. 10c). A large amount of moisture was transported toward southern Taiwan by the dominant west-southwesterly flow, consistent with Fig. 7b. The low at 1000 hPa had passed to the northeast of Taiwan and the front had moved further to the south on the Strait (Fig. 11e). Evident from the larger |∇θυ|, the front at this time had become even stronger (Fig. 11f). It produced strong lifting on the moist SW flow, leading to the heavy rainfall in southern Taiwan (Fig. 8e).
850-hPa mixing ratio (color shading; g kg−1), wind (vectors; m s−1), and geopotential height (contours; interval: 10 gpm) in the ensemble mean of the 64 members from domain 2 at (a) 0000 UTC 21 May, (b) 2100 UTC 21 May, and (c) 0600 UTC 22 May 2020. Gray color indicates the area below topography. Red box B in (a), which is the same as box 13 in Fig. 4a denotes the area for box average.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
1000-hPa (a),(c),(e) streamlines and horizontal divergence (color shading; unit: 1 × 10−4 s−1); and (b),(d),(f) winds (m s−1), virtual potential temperature (θυ; contours; interval: 1 K), and horizontal θυ gradient (|∇θυ|; color shading; unit: K 100 km−1) in the ensemble mean from domain 3 at (a),(b) 0000 UTC 21 May; (c),(d) 2100 UTC 21 May; and (e),(f) 0600 UTC 22 May 2020. The divergence and θυ gradient are computed in each individual member before an ensemble mean is taken. Box C in (a) denotes the area for front detection.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
b. Correlation analyses
In this subsection, the method developed in previous studies (e.g., Du and Chen 2018; Liu et al. 2020; Shen et al. 2020) is applied to calculate the correlation coefficient between precipitation in southern Taiwan and several atmospheric environmental fields at the 850-hPa level. The precipitation was first computed by an areal mean of the 6-h rainfall over land area in box A of Fig. 8a for all the 64 ensemble members during P1 and P2. Its correlation coefficients with the 850-hPa wind and moisture at a particular time were then evaluated at each grid point using the ensemble. Approximately 12 h before P1 (Fig. 12a), the SWV was located in southeastern China and had not yet affected the weather in Taiwan. At the starting time of P1 (Fig. 12d), the eastern portion of the SWV had just approached Taiwan, causing an increase of the westerly wind component over the Taiwan Strait. During these hours (Figs. 12a,d), an area of positive correlation coefficient between precipitation and the 850-hPa moisture gradually moved from the northern SCS to southwestern Taiwan. Because no significant weather system was dominating the wind field around Taiwan before P1, the correlations between rainfall and wind fields (either U or V) were all quite small (Figs. 12b,c). Only at the starting time of P1, did a small area of positive correlation coefficient between U and rainfall (Fig. 12e) and a larger area of negative correlation coefficient between V and rainfall (Fig. 12f) appear to the southwest of Taiwan. The latter suggests that if the upstream flow turned from SW to a more west-southwesterly direction (smaller V), there would be more rain in southern Taiwan.
Correlation coefficient (color shading; scale shown to the right) between the 850-hPa (a) mixing ratio, (b) x-component wind (U), and (c) y-component wind (V) at 12 h before P1 and the average 6-h rainfall over land area in box A of Fig. 8a in P1, calculated from the simulations of the 64 ensemble members. The 850-hPa wind (a full barb is 5 m s−1) and geopotential height (green contours; interval: 8 gpm) of the ensemble mean at 12 h before P1 are also shown. The numbers in the upper-right corner of each diagram are the maximum and minimum values of the correlation coefficients. (d)–(f) As in (a)–(c), but at the starting time of P1.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
During the 12 h before P2, the SWV was passing over the north side of Taiwan (Figs. 13a,d), and the environmental fields in the vicinity of Taiwan were changing to a state that was primarily dominated by the SWV. Affected by the increased geopotential height gradient, the SW wind speed over the southern Taiwan Strait and the northern SCS became stronger, and the wind direction was more westerly. The area of positive correlation between precipitation over southern Taiwan and the 850-hPa moisture gradually moved from southeastern China toward Taiwan (Figs. 13a,d), mostly following the west-southwesterly wind on the south side of the SWV. The areas where U (Figs. 13b,e) and V (Figs. 13c,f) were positively correlated with the rain in southern Taiwan also gradually moved toward Taiwan, but they originated from a more southwestward location such as the northern SCS. It is summarized that moisture over southeastern China is important for the rainfall over southern Taiwan in P2. The SW winds over the northern SCS are also important because when they are stronger, there will be more rainfall over southern Taiwan. The fact that the aforementioned correlation coefficients around Taiwan were overall larger in P2 than in P1 was mainly caused by the SWV, which enhanced winds, moisture fluxes, and rainfall in southern Taiwan when it moved closer.
c. Difference analyses
Using the 64 member runs, we further performed difference analyses of environmental fields between good-forecast members and bad-forecast members. Since most members in general underestimated rainfall in southern Taiwan, it should be fair to assume that members with more average rainfall are good members, while those with less rainfall are bad members. The seven ensemble members with the average rainfall over land area of box A in Fig. 8a ranked in the top 10% were selected as group A, and the seven members with rainfall ranked in the bottom 10% were group B. The analyses were carried out independently for both P1 and P2 periods such that there were four groups (P1A, P1B, P2A, and P2B) with different combinations of ensemble members. For P1A and P1B, the threshold of average precipitation was greater and smaller or equal to 74.8 and 44.3 mm (6 h)−1, respectively; for P2A and P2B, the threshold was 79.4 and 43.8 mm (6 h)−1, respectively. To examine the difference between the good and bad members, a composite average of rainfall and environmental fields was taken for each group.
During P1, the mean precipitation of P1A was mostly concentrated in the south, including the mountainous region and the low plains in southwestern Taiwan (Fig. 14a). In P1B, there was significantly less precipitation in southern Taiwan (Fig. 14b) but slightly more rainfall over the western plain area of central Taiwan than in P1A. It appears that the bad members (P1B) tended to produce rainfall that was not only relatively smaller, but also in the wrong location. At the beginning of P1 (Fig. 15a), when the SWV just reached the north side of Taiwan, large water vapor on the south side of the SWV was transported to the southern Taiwan Strait. The moisture difference between P1A and P1B (Fig. 15b) was small over the Strait. There was only a narrow area of small positive moisture difference extending from the northern SCS to Taiwan. Such difference indicates that moisture flux by the large-scale monsoonal flow from the SCS, rather than the SWV, was important in rainfall difference in P1. The 850-hPa geopotential height difference between P1A and P1B (Fig. 15c) was negative over the west coast of Taiwan and the Taiwan Strait, and positive in southern China. Based on the geostrophic balance, wind differences between P1A and P1B would be northerly around the southeast coast of China (see vectors in Figs. 15b or 15c). These results indicate that the SWV in P1A (Fig. 15a) shifted more to the south in its eastern portion (near Taiwan) and more to the north in its west portion (southern China) than that in P1B (not shown), resulting in a decrease in the southerly wind component on the SW winds in P1A. Winds over the ocean to the southwest of Taiwan in P1A (Fig. 15a) thus turned more west-southwesterly (or had a wind direction pointing slightly more to the east) than those in P1B. As a result, most MCSs moved over land in southern Taiwan in P1A, but in central Taiwan in P1B, leading to more precipitation in southern Taiwan in P1A.
Mean 6-h rainfall (unit: mm) of (a) the seven members with the most rainfall (P1A) and (b) the seven members with the least rainfall (P1B) over southern Taiwan in P1 (2100 UTC 21 May–0300 UTC 22 May 2020). The number in the lower-right corner is the maximum rainfall in the diagram. (c),(d) As in (a),(b), but for P2 (0600–1200 UTC 22 May 2020).
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
(a) Low-level-averaged (surface to 700 hPa) mixing ratio (color shading; g kg−1) and wind (vectors; m s−1), and the 850-hPa geopotential height (contours; interval: 10 gpm) from the ensemble mean of the seven members in P1A, and (b) the difference of low-level-averaged (surface to 700 hPa) mixing ratio (color shading; g kg−1) and wind (vectors; m s−1) between the ensemble means of P1A and P1B at 2100 UTC 21 May 2020. (c) As in (b), but color shading is for the 850-hPa geopotential height difference (gpm). The scale of color shading of each panel is denoted to the right of each row. (d)–(f) As in (a)–(c), but for P2 at 0600 UTC 22 May 2020.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
During P2, there was significantly more precipitation in P2A (Fig. 14c) than in P2B (Fig. 14d). The main rain axis in P2A was located in southwestern Taiwan, which was more consistent with the observation (Fig. 8d) than that in P2B which was too far south. At the beginning of P2, the SWV was passing over the north side of Taiwan (Fig. 15d), dominantly influencing the distribution of winds and moisture in the vicinity of Taiwan. The large water vapor area on the south side of the SWV was moving into southwestern Taiwan with the strong SW winds at this time. A large area of positive moisture difference between P2A and P2B extended from southeastern China to the west coast of Taiwan (Fig. 15e), suggesting that the more moisture upstream on the south side of the SWV, the more precipitation in southern Taiwan. As for the pressure difference between P2A and P2B, an area of positive geopotential height difference was extending from the ocean south of Taiwan to the ocean north-northeast of Taiwan (Fig. 15f), and a large area of negative geopotential height difference was found on the southwest side of the SWV covering most of southern and southeastern China, leading to a SW wind difference (Fig. 15f) over the Taiwan Strait and the northern SCS through the geostrophic balance. This result suggests that the SWV in P2A shifted more to the north on its east portion and more to the south on its west portion than in P2B (Fig. 15d), resulting in stronger SW winds and more precipitation in southern Taiwan in P2A. The SWV therefore plays an important role in enhancing the SW wind speed and transporting more water vapor from South China to Taiwan in P2.
d. Sensitivity analyses
The mei-yu front, as previously mentioned, plays an important role in lifting the moist air of the SW flow associated with the SWV to promote the formation of MCSs and rainfall. This subsection is aimed to determine the contribution of the front, the SWV, and other factors in a more quantitative way using the ensemble. To do so, we need to objectively locate the frontal position in the model simulations. An auto front-detection algorithm was thus developed in this study based on the method of large-scale front detection by Li et al. (2018). Our algorithm, revised for mesoscale fronts, includes the following steps:
-
In the domain of 21°–24.5°N, 118°–120°E (box C in Fig. 11a), find grid points that have horizontal divergence, ∇ ⋅ V < −7.1 × 10−4 s−1, and horizontal virtual potential temperature gradient, |∇θυ| > 0.14 K km−1 at 1000 hPa. These criteria are set for finding a front that meets both the dynamic and thermodynamic conditions.
-
For each grid point (i) in the x direction of box C, compute the mean latitude (Lati) of the grid points in the y direction that meet the above criteria. If no such points are found at a particular i, a missing value is given to its Lati. To ensure a relatively long and continuous front, two additional conditions are set: at least half of the Lati have to be non-missing value and the average Lati difference between two consecutive points has to be smaller than the distance of two grid points. If both of the criteria are met, a front is present in the domain and its latitude (LatF) is defined by averaging the above Lati, i.e.,
. If not, there will be no front.
The mean latitude of the fronts (
Figure 16a shows the boxplot of the frontal latitudes (LatF) for all the 64 ensemble members. Besides some outliers, the front was almost stationary during P1, with the majority of LatF being around 23°–23.5°N. The front started to move southward after P1 and moved faster during P2. The scatterplot of mean frontal latitudes
(a) Boxplot of frontal latitudes (LatF; ordinate) for the ensemble at 33–48 h into the simulation (2100 UTC 21 May–1200 UTC 22 May 2020), with the periods of P1 and P2 highlighted. The box extends from the first quartile to the third quartile (interquartile range), with a horizontal line denoting the median value. The upper and lower error bars show the data value that is 1.5 × interquartile range above and below the third and first quartile, respectively. Dots are outliers. (b) The scatterplot of mean latitudes of the front (
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
We further compute the sensitivity of the mean 6-h rainfall over land area in box A of Fig. 8a (R) to several variables (x) during P1 and P2, based on the ensemble sensitivity analysis method of Ancell and Hakim (2007) and Wang et al. (2021). The sensitivity that is defined as the ratio of the covariance of R and x to the variance of x represents the change in R when x increases by 1 unit in the ensemble. The greater the value, the more sensitive is R to x. The standard deviation (SD, or spread) of x in the ensemble is used to normalize the sensitivity such that the results (ΔR/ΔSD) are interpreted as the change in R (in mm) in response to an increase in x by one SD. In the following, the variables of the 850-hPa x-component wind (
During P1, the 850-hPa mixing ratio (
Normalized sensitivity (ΔR/ΔSD) of the mean 6-h rainfall over the land area in box A (R) to one standard deviation (SD) of variables (x) during P1. The variables (x) include the areal mean of the 850-hPa x-component wind (
SWV centers (color dots) of the 64 ensemble members in (a) P1 and (b) P2, with colors denoting the average rainfall [unit: mm (6 h)−1] over the land area in box A of Fig. 8a and sizes representing the average intensity of the SWV during the corresponding period. The color scale of rainfall is shown to the right, and the size scale is above the diagram. Four different dot sizes are used: the largest and the second largest denote the SWV with a geopotential height smaller than the mean, while the smallest and the second smallest larger than the mean, with the two in each pair separated by one SD from the mean. The values (unit: gpm) of one SD below the mean, the mean, and one SD above the mean are shown at the top. The mean locations of the SWV in the ensemble, the P1A(P2A) group, and the P1B(P2B) group are denoted in black, red, and blue crosses, respectively.
Citation: Monthly Weather Review 151, 7; 10.1175/MWR-D-22-0226.1
The SWV dominated the weather of Taiwan in P2 such that the sensitivity of the 850-hPa U and V were both largely positive during this period (Table 2). This result had two indications: first, the stronger the SW flow, the larger the rainfall in southern Taiwan; second, the stronger southerly flow could cause the front to situate at a more northward latitude, leading to more rainfall in southern Taiwan in P2 as discussed in Fig. 16. The sensitivity of the mean frontal latitude (
5. Summary and conclusions
During the first half of the mei-yu season (FHMY, 15–31 May) in 2020, a record-breaking extreme precipitation event occurred in Taiwan. The average accumulated precipitation of Taiwan’s 28 surface stations set a record of the maximum rainfall in FHMY in the history of Taiwan since 1979. The most extreme rainfall occurred during a 36-h period around 22 May when the average accumulated precipitation in Taiwan reached 135.9 mm per station, a climatological maximum of 36-h accumulated rainfall during FHMY in the previous 42 years. The composite results show that during FHMY in 2020, the PSH was significantly stronger than in the climatological mean, resulting in stronger monsoonal SW flows, which is consistent with the results of previous studies (Zeng et al. 2022; Takaya et al. 2020; Guo et al. 2021). Moreover, the water vapor conveyor belt extended from the northern Indo-China Peninsula to East Asia. The equivalent potential temperature gradient increased significantly to the north of Taiwan, leading to the increased frontal activity, SW flow generation, and MCS development near Taiwan. During the 36 h from 0000 UTC 21 May to 1200 UTC 22 May 2020, a southwest vortex (SWV) formed after strong westerly flow passed the southern side of the Tibetan Plateau and moved to the low-plain region of southwestern China. When the SWV moved eastward, the strong SW winds on its south side combined with the monsoonal SW flows and transported moisture continuously to Taiwan. Both the SW wind and moisture flux over the southern Taiwan Strait were extremely intense and ranked nearly the top in the climatology.
At the early times when the SWV was still located over southeastern China, the large-scale monsoonal SW flow dominated the water vapor flux to the southwest of Taiwan. However, since there was not much low-level convergence without the mei-yu front in the upstream flow, precipitation was small in southern Taiwan. In the first period of heavy rain in southern Taiwan (P1), the eastern portion of the SWV arrived at the north side of Taiwan. The west-southwesterly winds associated with the SWV and the SW winds of the large-scale monsoonal flow transported moisture toward the Taiwan Strait. The moisture-laden SW flow was lifted by the mei-yu front, leading to the development of MCSs over the ocean southwest to Taiwan. These MCSs moved over land and resulted in heavy rainfall in southern Taiwan. The correlation analyses show that moisture flux by the monsoonal flow from the SCS was more important than that by the SWV in determining rainfall. The SWV played a role in turning upstream flow from SW to a more west-southwesterly direction that influenced the precipitation area. When the southerly winds were stronger (weaker), the front tended to locate at a more northward (southward) latitude, resulting in smaller (larger) rainfall in southern Taiwan. It is found that locations of the front and the SWV were more important than their intensities in determining the rainfall during this period.
During the second period of heavy rain (P2), the SWV passed through the north side of Taiwan and the front moved southward on the Taiwan Strait. The SWV became the dominant weather system near Taiwan such that wind directions in the vicinity of Taiwan turned more westerly and the correlation between the environmental fields and precipitation was high. Since the main source of water vapor was from South China, upstream of westerly winds, when there was more water vapor in South China, there would be more precipitation in southern Taiwan. The SWV played a role in enhancing the SW wind speed during this period. When pressure was higher on the east coast of Taiwan and lower on the west side of the SWV, the SW wind speed was stronger and more water vapor was transported to Taiwan, leading to stronger precipitation. Similar to that in P1, frontal location, rather than intensity, was important in determining rainfall in southern Taiwan. Therefore, if southerly winds were stronger over the southern Taiwan Strait, the front would move southward with a slower speed and be located at a northward latitude that is closer to southern Taiwan, resulting in more rainfall there. As for the SWV, both its location and intensity were important in rainfall during this period.
To sum up, the 2020 mei-yu season is a special rainy season not only in Taiwan, but also in other regions of East Asia. The unusually strong PSH causes an early onset of strong SW flows and heavy rain in Taiwan during the FHMY. In the second half (1–15 June) of the mei-yu season, the mei-yu fronts already shift to a more northward latitude, and the rainy season in Taiwan ends early. Consequently, the mei-yu season starts relatively early in central China and Japan in early June and lasts until 31 July (Zeng et al. 2022; Wang et al. 2022). Our analyses conclude that the most critical factors leading to heavy rainfall in southern Taiwan are the strong 850-hPa SW winds (or LLJ) and moisture fluxes associated with the SW flow that is enhanced by the SWV. The distance of the front (SWV) to southern Taiwan is, by comparison the second (third), the frontal speed is the fourth, and the intensity of the SWV is the fifth most sensitive factors to the rainfall. However, it should be noted that there could be complex interactions among different factors; for example, it would be interesting to examine the sensitivities using only the ensemble members that simulated approximately right frontal locations. In this situation, the intensities of the front and the SWV may become more important in influencing rainfall. Previous studies (Tu et al. 2019, 2020; Chen et al. 2022) have indicated the important role that the marine boundary layer jet below 900 hPa over the SCS plays in the rainfall of Taiwan during mei-yu seasons. For the current case, the marine boundary layer jet may also help the synoptic LLJ in the moisture transport to the Taiwan area. This topic should be further investigated in a future study. Furthermore, although the role of terrain in enhancing rainfall, which has been well documented in the literature, is not discussed in this study, it should be noted that terrain is also another important factor in determining rainfall. Last, Chien and Chiu (2021) found that precipitation forecasts from models can be effectively improved if the dropsonde data from the upstream of the SW flow can be assimilated into the ETKF DA system. The correlation coefficient and sensitivity analyses of this paper can provide a good suggestion for the dropsonde observation collection.
Acknowledgments.
The data used in this study are obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) and the Central Weather Bureau of Taiwan (CWB). This research was supported by the Ministry of Science and Technology of Taiwan (Grants MOST 111-2111-M-003-004, MOST 110-2111-M-003-005, and MOST 110-2625-M-003-002). We thank the National Center for High-performance Computing (NCHC) for providing computational and storage resources and the three anonymous reviewers for their effort and constructive comments.
Data availability statement.
The ERA5 dataset is available at https://www.ecmwf.int/en/forecasts/dataset/ecmwf-reanalysis-v5. The datasets of surface weather stations, rain gauge stations, and satellite cloud imagery are available at https://dbar.pccu.edu.tw/member/ParameterSearch.aspx.
Two of the P1A members and one of the P1B members were defined as no front during P1. The divergence and |∇θυ| patterns show that there were several times during which the aforementioned step 2 did not pass because fronts were slightly shorter than the half-width of box C. The main cause was |∇θυ|, which was smaller than the threshold at half of the grid points. The convergence was actually not small. Therefore, the heavy rain of these P1A members may be more related to the dynamic than the thermodynamic effects of the front.
REFERENCES
Akaeda, K., J. Reisner, and D. Parsons, 1995: The role of mesoscale and topographically induced circulations in initiating a flash flood observed during the TAMEX project. Mon. Wea. Rev., 123, 1720–1739, https://doi.org/10.1175/1520-0493(1995)123<1720:TROMAT>2.0.CO;2.
Ancell, B., and G. J. Hakim, 2007: Comparing adjoint- and ensemble-sensitivity analysis with applications to observation targeting. Mon. Wea. Rev., 135, 4117–4134, https://doi.org/10.1175/2007MWR1904.1.
Araki, K., T. Kato, Y. Hirockawa, and W. Mashiko, 2021: Characteristics of atmospheric environments of quasi-stationary convective bands in Kyushu, Japan during the July 2020 heavy rainfall event. SOLA, 17, 8–15, https://doi.org/10.2151/sola.2021-002.
Barker, D., and Coauthors, 2012: The Weather Research and Forecasting Model’s Community Variational/Ensemble Data Assimilation System: WRFDA. Bull. Amer. Meteor. Soc., 93, 831–843, https://doi.org/10.1175/BAMS-D-11-00167.1.
Bishop, C. H., B. J. Etherton, and S. J. Majumdar, 2001: Adaptive sampling with the ensemble transform Kalman filter. Part I: Theoretical aspects. Mon. Wea. Rev., 129, 420–436, https://doi.org/10.1175/1520-0493(2001)129<0420:ASWTET>2.0.CO;2.
Chen, C.-S., W.-C. Chen, Y.-L. Chen, P.-L. Lin, and H.-C. Lai, 2005: Investigation of orographic effects on two heavy rainfall events over southwestern Taiwan during the Mei-Yu season. Atmos. Res., 73, 101–130, https://doi.org/10.1016/j.atmosres.2004.07.005.
Chen, G. T.-J., 1983: Observational aspects of the Mei-Yu phenomenon in subtropical China. J. Meteor. Soc. Japan, 61, 306–312, https://doi.org/10.2151/jmsj1965.61.2_306.
Chen, G. T.-J., 1992: Mesoscale features observed in the Taiwan Mei-Yu season. J. Meteor. Soc. Japan, 70, 497–516, https://doi.org/10.2151/jmsj1965.70.1B_497.
Chen, G. T.-J., and C.-C. Yu, 1988: Study of low-level jet and extremely heavy rainfall over northern Taiwan in the Mei-Yu season. Mon. Wea. Rev., 116, 884–891, https://doi.org/10.1175/1520-0493(1988)116<0884:SOLLJA>2.0.CO;2.
Chen, G. T.-J., and C.-Y. Liang, 1992: A midlevel vortex observed in the Taiwan area mesoscale experiment (TAMEX). J. Meteor. Soc. Japan, 70, 25–41, https://doi.org/10.2151/jmsj1965.70.1_25.
Chen, G. T.-J., C.-C. Wang, and L.-F. Lin, 2006: A diagnostic study of a retreating Mei-Yu front and the accompanying low-level jet formation and intensification. Mon. Wea. Rev., 134, 874–896, https://doi.org/10.1175/MWR3099.1.
Chen, G. T.-J., C.-C. Wang, and S.-W. Chang, 2008: A diagnostic case study of Mei-Yu frontogenesis and development of wavelike frontal disturbances in the subtropical environment. Mon. Wea. Rev., 136, 41–61, https://doi.org/10.1175/2007MWR1966.1.
Chen, X. A., and Y.-L. Chen, 1995: Development of low-level jets during TAMEX. Mon. Wea. Rev., 123, 1695–1719, https://doi.org/10.1175/1520-0493(1995)123<1695:DOLLJD>2.0.CO;2.
Chen, Y.-L., and N. B.-F. Hui, 1990: Analysis of a shallow front during the Taiwan Area Mesoscale Experiment. Mon. Wea. Rev., 118, 2649–2667, https://doi.org/10.1175/1520-0493(1990)118<2649:AOASFD>2.0.CO;2.
Chen, Y.-L., and N. B.-F. Hui, 1992: Analysis of a relatively dry front during the Taiwan area mesoscale experiment. Mon. Wea. Rev., 120, 2442–2468, https://doi.org/10.1175/1520-0493(1992)120<2442:AOARDF>2.0.CO;2.
Chen, Y.-L., Y.-X. Zhang, and N. B.-F. Hui, 1989: Analysis of a surface front during the early summer rainy season over Taiwan. Mon. Wea. Rev., 117, 909–931, https://doi.org/10.1175/1520-0493(1989)117<0909:AOASFD>2.0.CO;2.
Chen, Y.-L., C.-C. Tu, F. Hsiao, C.-S. Chen, P.-L. Lin, and P.-H. Lin, 2022: An overview of low-level jets (LLJs) and their roles in heavy rainfall over the Taiwan area during the early summer rainy season. Meteorology, 1, 64–112, https://doi.org/10.3390/meteorology1010006.
Chien, F.-C., 2015: The role of southwesterly flow in MCS formation during a heavy rain event in Taiwan on 12–13 June 2005. Terr. Atmos. Oceanic Sci., 26, 411–429, https://doi.org/10.3319/TAO.2015.03.11.01(A).
Chien, F.-C., and Y.-C. Chiu, 2019: A composite study of southwesterly flows and rainfall in Taiwan. J. Meteor. Soc. Japan, 97, 1023–1040, https://doi.org/10.2151/jmsj.2019-057.
Chien, F.-C., and Y.-C. Chiu, 2021: Assessing the impact of dropsonde data on rain forecasts in Taiwan with observing system simulation experiments. Atmosphere, 12, 1672, https://doi.org/10.3390/atmos12121672.
Chien, F.-C., Y.-C. Chiu, and C.-H. Tsou, 2021: A climatological study of southwesterly flows and heavy precipitation in Taiwan during Mei-Yu seasons from 1979 to 2018. J. Meteor. Soc. Japan, 99, 913–931, https://doi.org/10.2151/jmsj.2021-044.
Chou, L. C., C.-P. Chang, and R. T. Williams, 1990: A numerical simulation of the Mei-Yu front and the associated low-level jet. Mon. Wea. Rev., 118, 1408–1428, https://doi.org/10.1175/1520-0493(1990)118<1408:ANSOTM>2.0.CO;2.
Ding, Y., 1992: Summer monsoon rainfalls in China. J. Meteor. Soc. Japan, 70, 373–396, https://doi.org/10.2151/jmsj1965.70.1B_373.
Ding, Y., and J. C. L. Chan, 2005: The East Asian summer monsoon: An overview. Meteor. Atmos. Phys., 89, 117–142, https://doi.org/10.1007/s00703-005-0125-z.
Du, Y., and G. Chen, 2018: Heavy rainfall associated with double low-level jets over southern China. Part I: Ensemble-based analysis. Mon. Wea. Rev., 146, 3827–3844, https://doi.org/10.1175/MWR-D-18-0101.1.
Feng, X., C. Liu, G. Fan, X. Liu, and C. Feng, 2016: Climatology and structures of southwest vortices in the NCEP Climate Forecast System reanalysis. J. Climate, 29, 7675–7701, https://doi.org/10.1175/JCLI-D-15-0813.1.
Guo, Y., R. Zhang, Z. Wen, J. Li, C. Zhang, and Z. Zhou, 2021: Understanding the role of SST anomaly in extreme rainfall of 2020 Meiyu season from an interdecadal perspective. Sci. China Earth Sci., 64, 1619–1632, https://doi.org/10.1007/s11430-020-9762-0.
Hersbach, H., 2016: The ERA5 atmospheric reanalysis. 2016 Fall Meeting, San Francisco, CA, Amer. Geophys. Union, Abstract NG33D–01.
Hirockawa, Y., T. Kato, K. Araki, and W. Mashiko, 2020: Characteristics of an extreme rainfall event in Kyushu district, southwestern Japan in early July 2020. SOLA, 16, 265–270, https://doi.org/10.2151/sola.2020-044.
Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, https://doi.org/10.1175/MWR3199.1.
Horinouchi, T., Y. Kosaka, H. Nakamigawa, H. Nakamura, N. Fujikawa, and Y. N. Takayabu, 2021: Moisture supply, jet, and silk-road wave train associated with the prolonged heavy rainfall in Kyushu, Japan in early July 2020. SOLA, 17B, 1–8, https://doi.org/10.2151/sola.2021-019.
Hsu, W.-R., and W.-Y. Sun, 1994: A numerical study of a low-level jet and its accompanying secondary circulation in a mei-yu system. Mon. Wea. Rev., 122, 324–340, https://doi.org/10.1175/1520-0493(1994)122<0324:ANSOAL>2.0.CO;2.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, https://doi.org/10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Kuo, Y.-H., and G. T.-J. Chen, 1990: The Taiwan Area Mesoscale Experiment (TAMEX): An overview. Bull. Amer. Meteor. Soc., 71, 488–503, https://doi.org/10.1175/1520-0477(1990)071<0488:TTAMEA>2.0.CO;2.
Kuo, Y.-H., L. Cheng, and J.-W. Bao, 1988: Numerical simulation of the 1981 Sichuan flood. Part I: Evolution of the mesoscale southwest vortex. Mon. Wea. Rev., 116, 2481–2504, https://doi.org/10.1175/1520-0493(1988)116<2481:NSOTSF>2.0.CO;2.
Lau, K.-M., G. J. Yang, and S. H. Shen, 1988: Seasonal and intraseasonal climatology of summer monsoon rainfall over East Asia. Mon. Wea. Rev., 116, 18–37, https://doi.org/10.1175/1520-0493(1988)116<0018:SAICOS>2.0.CO;2.
Li, J., Y.-L. Chen, and W.-C. Lee, 1997: Analysis of a heavy rainfall event during TAMEX. Mon. Wea. Rev., 125, 1060–1082, https://doi.org/10.1175/1520-0493(1997)125<1060:AOAHRE>2.0.CO;2.
Li, Y., Y. Deng, S. Yang, and H. Zhang, 2018: Multi-scale temporospatial variability of the East Asian Meiyu-Baiu fronts: Characterization with a suite of new objective indices. Climate Dyn., 51, 1659–1670, https://doi.org/10.1007/s00382-017-3975-4.
Lin, Y.-J., R. W. Pasken, and H.-W. Chang, 1992: The structure of a subtropical prefrontal convective rainband. Part I: Mesoscale kinematic structure determined from dual-Doppler measurements. Mon. Wea. Rev., 120, 1816–1836, https://doi.org/10.1175/1520-0493(1992)120<1816:TSOASP>2.0.CO;2.
Liu, X., Y. Luo, L. Huang, D. L. Zhang, and Z. Guan, 2020: Roles of double low‐level jets in the generation of coexisting inland and coastal heavy rainfall over south China during the presummer rainy season. J. Geophys. Res. Atmos., 125, e2020JD032890, https://doi.org/10.1029/2020JD032890.
Lu, M.-M., C.-H. Sui, J.-R. Sun, and P.-H. Lin, 2020: Influences of subseasonal to interannual oscillations on the SCS summer monsoon onset in 2018. Terr. Atmos. Oceanic Sci., 31, 197–209, https://doi.org/10.3319/TAO.2020.02.25.01.
Ralph, F. M., J. J. Rutz, J. M. Cordeira, M. Dettinger, M. Anderson, D. Reynolds, L. J. Schick, and C. Smallcomb, 2019: A scale to characterize the strength and impacts of atmospheric rivers. Bull. Amer. Meteor. Soc., 100, 269–289, https://doi.org/10.1175/BAMS-D-18-0023.1.
Ray, P. S., A. Robinson, and Y. Lin, 1991: Radar analysis of a TAMEX frontal system. Mon. Wea. Rev., 119, 2519–2539, https://doi.org/10.1175/1520-0493(1991)119<2519:RAOATF>2.0.CO;2.
Shen, Y., Y. Du, and G. Chen, 2020: Ensemble sensitivity analysis of heavy rainfall associated with three MCSs coexisting over southern China. J. Geophys. Res. Atmos., 125, e2019JD031266, https://doi.org/10.1029/2019JD031266.
Shu, Y., J. Sun, and J. Chenlu, 2022: A 10-year climatology of midlevel mesoscale vortices in China. J. Appl. Meteor. Climatol., 61, 309–328, https://doi.org/10.1175/JAMC-D-21-0095.1.
Skamarock, W. C., and Coauthors, 2019: A description of the Advanced Research WRF Model version 4. NCAR Tech. Note NCAR/TN-556+STR, 145 pp., https://doi.org/10.5065/1dfh-6p97.
Tai, S.-L., Y.-C. Liou, S.-F. Chang, and J. Sun, 2020: The heavy rainfall mechanism revealed by a terrain-resolving 4DVar data assimilation system—A case study. Mon. Wea. Rev., 148, 2307–2330, https://doi.org/10.1175/MWR-D-19-0244.1.
Takaya, Y., I. Ishikawa, C. Kobayashi, H. Endo, and T. Ose, 2020: Enhanced Meiyu‐Baiu rainfall in early summer 2020: Aftermath of the 2019 super IOD event. Geophys. Res. Lett., 47, e2020GL090671, https://doi.org/10.1029/2020GL090671.
Tao, W.-K., J. Simpson, and M. McCumber, 1989: An ice-water saturation adjustment. Mon. Wea. Rev., 117, 231–235, https://doi.org/10.1175/1520-0493(1989)117<0231:AIWSA>2.0.CO;2.
Teng, J.-H., C.-S. Chen, T.-C. C. Wang, and Y.-L. Chen, 2000: Orographic effects on a squall line system over Taiwan. Mon. Wea. Rev., 128, 1123–1138, https://doi.org/10.1175/1520-0493(2000)128<1123:OEOASL>2.0.CO;2.
Thomas, C. M., and D. M. Schultz, 2019: What are the best thermodynamic quantity and function to define a front in gridded model output? Bull. Amer. Meteor. Soc., 100, 873–895, https://doi.org/10.1175/BAMS-D-18-0137.1.
Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117, 1779–1800, https://doi.org/10.1175/1520-0493(1989)117<1779:ACMFSF>2.0.CO;2.
Trier, S. B., D. B. Parsons, and T. J. Matejka, 1990: Observations of a subtropical cold front in a region of complex terrain. Mon. Wea. Rev., 118, 2449–2470, https://doi.org/10.1175/1520-0493(1990)118<2449:OOASCF>2.0.CO;2.
Tu, C.-C., Y.-L. Chen, C.-S. Chen, P.-L. Lin, and P.-H. Lin, 2014: A comparison of two heavy rainfall events during the Terrain-Influenced Monsoon Rainfall Experiment (TiMREX) 2008. Mon. Wea. Rev., 142, 2436–2463, https://doi.org/10.1175/MWR-D-13-00293.1.
Tu, C.-C., Y.-L. Chen, P.-L. Lin, and Y. Du, 2019: Characteristics of the marine boundary layer jet over the South China Sea during the early summer rainy season of Taiwan. Mon. Wea. Rev., 147, 457–475, https://doi.org/10.1175/MWR-D-18-0230.1.
Tu, C.-C., Y.-L. Chen, P.-L. Lin, and P.-H. Lin, 2020: The relationship between the boundary layer moisture transport from the South China Sea and heavy rainfall over Taiwan. Terr. Atmos. Oceanic Sci., 31, 159–176, https://doi.org/10.3319/TAO.2019.07.01.01.
Tu, C.-C., Y.-L. Chen, P.-L. Lin, and M.-Q. Huang, 2022: Analysis and simulations of a heavy rainfall event associated with the passage of a shallow front over Northern Taiwan on 2 June 2017. Mon. Wea. Rev., 150, 505–528, https://doi.org/10.1175/MWR-D-21-0113.1.
Volonté, A., A. G. Turner, R. Schiemann, P. L. Vidale, and N. P. Klingaman, 2022: Characterising the interaction of tropical and extratropical air masses controlling East Asian summer monsoon progression using a novel frontal detection approach. Wea. Climate Dyn., 3, 575–599, https://doi.org/10.5194/wcd-3-575-2022.
Wang, B., 2006: The Asian Monsoon. Springer, 787 pp.
Wang, C.-C., J. C.-S. Hsu, G. T.-J. Chen, and D.-I. Lee, 2014: A study of two propagating heavy-rainfall episodes near Taiwan during SoWMEX/TiMREX IOP-8 in June 2008. Part I: Synoptic evolution, episode propagation, and model control simulation. Mon. Wea. Rev., 142, 2619–2643, https://doi.org/10.1175/MWR-D-13-00331.1.
Wang, C.-C., M.-S. Li, C.-S. Chang, P.-Y. Chuang, S.-H. Chen, and K. Tsuboki, 2021: Ensemble-based sensitivity analysis and predictability of an extreme rainfall event over northern Taiwan in the Mei-Yu season: The 2 June 2017 case. Atmos. Res., 259, 105684, https://doi.org/10.1016/j.atmosres.2021.105684.
Wang, X., T. M. Hamill, J. S. Whitaker, and C. H. Bishop, 2007: A comparison of hybrid ensemble transform Kalman filter-optimum interpolation and ensemble square root filter analysis schemes. Mon. Wea. Rev., 135, 1055–1076, https://doi.org/10.1175/MWR3307.1.
Wang, X., R. Zhou, Y. Deng, C. Cui, Y. Hu, J. Wang, and H. Liu, 2022: Symbiotic relationship between mei-yu rainfall and the morphology of mei-yu front. J. Hydrometeor., 23, 87–100, https://doi.org/10.1175/JHM-D-21-0068.1.
Xu, W., E. J. Zipser, Y.-L. Chen, C. Liu, Y.-C. Liou, W.-C. Lee, and B. J.-D. Jou, 2012: An orography-associated extreme rainfall event during TiMREX: Initiation, storm evolution, and maintenance. Mon. Wea. Rev., 140, 2555–2574, https://doi.org/10.1175/MWR-D-11-00208.1.
Zeng, W., G. Chen, L. Bai, Q. Liu, and Z. Wen, 2022: Multiscale processes of heavy rainfall over East Asia in summer 2020: Diurnal cycle in response to synoptic disturbances. Mon. Wea. Rev., 150, 1355–1376, https://doi.org/10.1175/MWR-D-21-0308.1.
Zhang, Q.-H., K.-H. Lau, Y.-H. Kuo, and S.-J. Chen, 2003: A numerical study of a mesoscale convective system over the Taiwan strait. Mon. Wea. Rev., 131, 1150–1170, https://doi.org/10.1175/1520-0493(2003)131<1150:ANSOAM>2.0.CO;2.
