Climatic characteristics and main weather patterns of extreme precipitation in the middle Yangtze River valley Open Access

In recent decades, the annual maximum of global land precipitation has increased with the increase of global mean surface temperature due to the impact of global warming (Westra et al. 2013). The IPCC Fifth Assessment Report also indicates that the number of heavy precipitation events on land has increased over more areas than it has decreased since about 1950 (Hartmann et al. 2013). It is generally accepted that the characteristics of future climate change will be an increase in extreme precipitation events due to global warming (Zhai et al. 2005; Zhang et al. 2013; Li et al. 2018a; Dong et al. 2020). As extreme precipitation can cause waterlogging of land, river overflow, farmland destruction, and house collapse, it can also lead to secondary disasters such as flash floods, landslides, debris flows, and waterlogging, often resulting in major casualties and property losses and causing greater social impact. Improving the understanding and prediction of extreme precipitation events will help to reduce these socioeconomic losses, so there are more and more studies on extreme precipitation events (Donat et al. 2013; Chen & Sun 2017; Sillmann et al. 2017; Li et al. 2018b).

The analysis of the climatic characteristics of extreme precipitation is helpful to understand the spatial and temporal distribution and change trend of extreme precipitation, and improve the prediction ability of such events. Studies have shown that extreme precipitation has obvious interannual and interdecadal variability characteristics (Tang et al. 2006; Li et al. 2023). Zhong et al. (2020) find that the total amount and frequency of extreme precipitation in summer are concentrated in the Yangtze River Basin and south China, and the EOF1 decomposition of extreme precipitation shows the interannual oscillation characteristics of reverse spatial distribution and the significant interannual characteristics. Wang & Qian (2009) studied the characteristics of the spatial-temporal distributions of their frequencies, intensities, and the linear trend of extreme precipitation events in China. The results show that the frequency of extreme precipitation events is higher in the regions south of 35°N than in other parts of China, especially in the mid-lower reaches and the southern parts of the Yangtze River. However, these studies are all aimed at large-scale regional areas and cannot fully reflect the subtle characteristics of the distribution, and the study of extreme precipitation from a smaller spatial scale is an important supplement to large-scale regional research.

Northern California Classification is an effective method to study extreme precipitation and related large-scale circulation changes in synoptic climatology (Liu et al. 2015). Extreme precipitation processes generally have typical weather conditions. By classifying the influence systems of extreme precipitation processes, it is helpful for us to understand the weather scale and mesoscale mechanisms that lead to extreme precipitation. There are two kinds of classification methods: subjective classification and objective classification. The disadvantage of subjective classification is that it relies too much on the subjective judgment of the researchers. Due to the different statistical samples, it is often difficult to unify the classification results of different researchers. The mathematical objective classification based on the basic elements may lead to the clustering classification with unclear weather significance. The two methods have their own advantages and disadvantages. In terms of subjective classification, Maddox et al. (1979) investigated 151 flash flood events in the United States and their temporal and spatial characteristics. According to the surface and upper air situation, these flash floods are divided into four weather types: weather scale forcing type, frontal type, medium-high pressure type, and western type. This classification is still widely used by the US weather forecast department. According to the weather conditions at the time of occurrence, Luo et al. (2016) divided the extreme hourly precipitation in China into four types: tropical cyclone type, frontal type, vortex/shear line type, and weak weather forcing type. According to the 500 hPa circulation situation, Xiao et al. (2017) divided extreme rainstorms in the Sichuan Basin into two types: east high and west low type and two high shear type. At the same time, the conceptual model of extreme rainstorms with different circulation types and different rainstorm centers was summarized. Salvador et al. (2022) used multivariate analysis methods (the diagnostic quantities used were sea level pressure, 850 hPa temperature, and 500 hPa geopotential height) to classify the weather systems that caused floods in the Mediterranean coast of Spain from 1960 to 2015 in 12 weather patterns, of which two models produced 59% of flood events.

In terms of objective classification, Hu et al. (2015) used the multivariate empirical orthogonal function (MV-EOF) technique to divide the dominant circulation models of the extreme precipitation process in the Sichuan Basin of China in 2013 into two types. Zhao et al. (2017) used the minimum variance clustering method to classify the extreme precipitation events in the continental United States from 1950 to 2005 and identified six typical extreme precipitation modes in the warm season and five modes in the cold season of the Northern Hemisphere. Based on the classification results, the large-scale weather models of each mode were derived, and the potential of climate models to accurately characterize different extreme precipitation modes was evaluated. Yang et al. (2019) used the k-means clustering method to divide the extreme precipitation events in the North China Plain from 1979 to 2016 into the northern mode and the central-southern mode and revealed the synergistic variation characteristics of the two modes’ influence systems. Zhou et al. (2020) used the hierarchical clustering method to divide the persistent extreme rainstorm events in North China into four categories according to the circulation background: meridional type, zonal type, weakened landing tropical cyclone type, and early summer type. These studies have a good guiding role in operational forecasting.

The middle Yangtze River Valley (MYRV) is one of the most important agricultural areas and densely populated areas in China (Lu 2000), with rapid economic development. It is also a concentrated area of flood disasters and is highly sensitive to catastrophic events. Under the climate background of the increasing intensity and frequency of extreme heavy precipitation in the future (Ren et al. 2014; Zhang et al. 2015a; Ke & Guan 2017), due to the complexity and regional differences of the extreme precipitation influence system, systematically summarizing the climatic characteristics and main weather patterns of extreme precipitation in this region will help to improve the prediction ability of such events and reduce the social and economic losses caused by extreme precipitation.

The data used in this article include: (1) Daily precipitation data of 315 national stations in the MYRV from 1981 to 2020 for 40 years. The daily precipitation period is 08:00–08:00 (Beijing time, the same below); (2) ERA5 data (Hersbach et al. 2020) provided by the European Centre for Medium-Range Weather Forecasts, including 500 hPa height field and 850 hPa wind field, with a spatial resolution of 0.25° × 0.25°.

In the aforementioned formula, m is the serial number of x_n. According to the aforementioned method, the 99th percentile value of each station from 1981 to 2020 is calculated, and the daily extreme precipitation threshold is the average of the 99th percentile values of 40 years (Zhai & Pan 2003). Finally, the daily precipitation of each station from 1981 to 2020 is counted. If the daily precipitation of a station is greater than or equal to the average daily extreme precipitation threshold of the station over the years, the day is counted as an extreme precipitation day.

The definition of regional extreme precipitation days adopts the standard of Wang et al. (2018). Wang et al. (2018) stipulate that if the daily rainfall of more than 5% of the stations in the study area reaches the rainfall intensity, it is defined as a regional rainfall day. Therefore, if more than 5% of the 315 national stations in the MYRV (≥16 stations) have extreme precipitation on a certain day, that is to say, at least 16 stations reached the standard of extreme precipitation days on the same day, and then this day is considered as a regional extreme precipitation day.

To obtain the co-variation characteristics of the elements of the atmospheric circulation systems of extreme precipitation in the MYRV, the MV-EOF (Wang 1992; Wang et al. 2008) was used to conduct spatial-temporal decomposition of multiple meteorological element fields. This method can obtain the spatial distribution of different variable fields with the same time coefficient. In this article, 850 hPa wind field (U and V components), 500 hPa height field, and surface precipitation are selected for MV-EOF decomposition. Due to the large difference in the magnitude of different elements, all elements are standardized before calculation. The analysis time is 08:00 of the regional extreme precipitation day, and the key area of the study is the MYRV (104–120°E, 24–35°N, Figure 1). The spatial modes separated after MV-EOF calculation represent the spatial distribution and synergistic variation characteristics of each physical quantity.

Spectral clustering is used in this article. Compared with the K-means method, spectral clustering has stronger adaptability to data distribution, better clustering effect, and less computation (Luxburg 2007). It performs traditional clustering on affinity matrices rather than raw data and is rarely used in the field of atmospheric science (Tang et al. 2021). In this study, a Python machine learning package including spectral clustering (Pedregosa et al. 2011) was used to cluster the 500 hPa height field, 850 hPa wind field (U and V wind components), and surface precipitation field in the MYRV. The physical quantities on each grid point are reduced to a pure time series and a normalized time series. Finally, the classification clustering sequence corresponding to the time series is obtained for weather model analysis.

During the past 40 years (1981–2020), a total of 15,707 extreme precipitation days occurred at 315 stations in the MYRV, with an average of approximately 393 stations per year and an average of 50 times per station. The interannual variation of extreme precipitation days is significant, with a minimum of 222 times (2001) and a maximum of 652 times (1998). Table 1 shows that extreme precipitation days can occur in any month of the year, but the frequency of occurrence in each month varies greatly, which has obvious inter month variation characteristics. Extreme precipitation days mainly occur from April to October, especially from May to August, accounting for more than 80% of the total number, of which June is the most, followed by July, accounting for 29.5 and 25.2%, respectively, and winter (December to February) is the least.

During the past 40 years, there have been 215 regional extreme precipitation days in the MYRV, and the average annual number of extreme precipitation days is 5.37, but the annual distribution is uneven. There are 4 years that occur 10 or more regional extreme precipitation days. Among them, there are 12 times in 1996. It only occurred once in 2015. On June 19, 2010, 70 stations in the MYRV met the standard of extreme precipitation day, which is the most reaching the standard. From the monthly frequency of regional extreme precipitation days (Table 2), it can be seen that except for winter (December to February), regional extreme precipitation days occur in other months, but mainly between May to August, of which June is the most, accounting for about 40%, followed by July, accounting for more than one-fourth.

The weather characteristics and occurrence mechanisms of regional extreme precipitation under different circulation configurations are different. To better understand the causes of extreme precipitation in the MYRV, this section uses the MV-EOF method and the spectral clustering method to objectively classify and compare the situation field configuration of 215 regional extreme precipitation days from 1981 to 2020.

The 500 hPa height field, 850 hPa wind field (U and V wind components), and surface precipitation field of 215 regional extreme precipitation days at 08:00 are decomposed by MV-EOF. Table 3 shows that only the first three modes of the first five modes passed the significance test (North et al. 1982), and their variance contribution rates were 20.9, 12.4, and 8.9%, respectively. The cumulative variance contribution rate was 42.3%, indicating that the first three modes were independent of each other and could be significantly distinguished from other modes.

The second mode configuration is directly affected by abnormal convergence and divergence (Figure 4(b)). When the time series is positive, the MYRV is controlled by the 500 hPa deep upper trough, and the 850 hPa is affected by the cold shear line formed by the intersection of abnormal northeast wind and abnormal southwest wind. The southern part of the MYRV is located in the convergence area near the cold shear line, corresponding to the positive anomaly area of precipitation, and the northern part of the MYRV is the negative anomaly area of precipitation. When the time series is negative, the southeast of the MYRV is an anticyclonic divergence area, which is not conducive to precipitation, corresponding to the negative anomaly area of precipitation. The northwest of the MYRV is affected by the consistent anomalous southwest wind, and there are obvious strong wind core area and cyclonic shear convergence area, corresponding to the positive anomaly area of precipitation.

In addition, Figure 4 also shows that the first two modes reflect the characteristics of the north-south antiphase distribution in the spatial distribution of precipitation, which is consistent with the north-south antiphase distribution of extreme precipitation magnitude and frequency presented (Figure 3).

It should be noted that since the variables are standardized before the MV-EOF decomposition, the anomaly fields of each element in Figure 4 only represent the deviation degree of the average state of the background fields and do not fully represent the configuration of the circulation situation. The larger the absolute value of the time coefficient of the eigenvectors, the more typical the distribution pattern of this type (Wei 2007). Therefore, to analyze the characteristics of the atmospheric circulation situation corresponding to the first two modes more intuitively, the background fields of the typical cases (Figure 5) with the standard deviation of the time series after decomposition greater than 1 (positive anomaly) and less than −1 (negative anomaly) are synthesized and analyzed, and the circulation configuration that causes extreme precipitation in the MYRV are further discussed. According to the aforementioned criteria, there are nine positive anomalies in extreme precipitation days and 20 negative anomalies in the first mode, and 29 positive anomalies and 35 negative anomalies in the second mode (Figure 5).

Figure 6(a) shows the composite circulation based on the typical case of the positive anomaly of the first mode, which can be summarized as a southwest vortex pattern. Although the southwest vortex is generated in the western Sichuan Plateau, its development and eastward movement will cause heavy rain and flood disasters in the vast areas of eastern China (Feng et al. 2016; Li & Chen 2018), and it is a very important weather system leading to extreme precipitation in the MYRV. At the low level, the distinctive feature of this type is the southwest vortex and shear lines in the northeast and southwest of the southwest vortex. The deep southwest warm and humid airflow on the south side of the warm shear line extends above 700 hPa, and generally has a low-level jet (LLJ). The middle layer is a deep trough, and there is a strong positive vorticity advection in front of the trough. Heavy rainfall is usually located near the warm shear line and its south side, distributing in quasi-east-west direction along the warm shear line. The southwest warm and humid airflow with abundant water vapor content and strong thermal instability provides a favorable unstable environment for the occurrence of extreme precipitation. The large-scale dynamic forcing in front of the middle-level trough and the cyclonic shear convergence on the left side of the low-level strong southwest airflow is conducive to maintaining a long-term upward movement in a large area.

Figure 6(b) shows the composite circulation based on the typical case of the negative anomaly of the first mode, which can be summarized as typhoon type. Although the MYRV is located inland, landing typhoons still have a greater impact on the region. For example, the 200,604 ‘Bili’ and 200,709 ‘Sepat,’ which had a greater impact at the beginning of this century, caused serious floods in Jiangxi and Hunan (Ye et al. 2009; Cheng et al. 2013). Typhoon type extreme precipitation mainly occurs in the midsummer season, which is the strongest period of the Western Pacific Subtropical High (WPSH), with stronger intensity and northward location (Figure 6(b)). The typhoon's low-pressure circulation is located on the southwest side of the WPSH, which is conducive to the northwestward movement of the typhoon. The low-pressure circulation is affected by two lower-level jets, which are the southerly jet on the east side of the low-pressure circulation and the northerly jet on the west side of the low-pressure circulation. The heavy rainfall occurs in two locations, one is near the inverted trough on the north side of the low-pressure circulation, and the other is on the southwest side of the low-pressure circulation, especially on the southwest side of the low-pressure circulation. Affected by the topography of the southern region (Figure 1), the southwest is the main heavy rainfall area of this type (Yao et al. 2007; Gao et al. 2009).

Figure 6(c) shows the circulation situation synthesized by the typical cases based on the positive anomaly of the second mode of MV-EOF, which can be summarized as cold trough shear line type. This type may include two situations, the first of which is shown in Figure 6(c). The influencing systems are the 500 hPa trough and the 850 hPa cold shear line. There may be a weak cyclonic circulation or southwest vortex upstream, but the cyclonic circulation stays in place and does not move eastward to affect the MYRV. The warm shear line on the east side of the cyclone is weak, and the extreme precipitation is mainly affected by the cold shear line between the southwest wind and the northeast wind. In the second case, when the southwest vortex moves out of the MYRV (because the studied area of MV-EOF decomposition is limited to 104–120°E and 24–35°N, the southwest vortex cannot be seen in Figure 6(c)), the MYRV is affected by the high-altitude cold trough and cold shear line behind the southwest vortex. During the period of extreme precipitation, the southwest vortex does not directly affect the region. The direct impact of extreme precipitation is still the 500 hPa high-altitude trough and 850 hPa cold shear line. At the low level, there is generally a LLJ on the south side of the shear line. Extreme precipitation occurs in the shear line and its south side of the wind direction shear and wind speed convergence area (usually located south of the Yangtze River, which is slightly more southerly than the location of the southwest vortex type).

Figure 6(d) shows the circulation situation synthesized by the typical cases based on the negative anomaly of the second mode of MV-EOF. From the synthetic circulation situation, Figure 6(d) is basically consistent with Figure 6(a) and belongs to the southwest vortex type.

It can be seen from the analysis of the previous section that the extreme precipitation situation field in the MYRV after MV-EOF decomposition can be divided into three types. The three types of situation fields well reflect the local changes and collaborative changes of atmospheric circulation in the MYRV. In this section, the spectral clustering method is used to calculate the three classifications to verify the classification results of MV-EOF.

Comparing the three types of extreme precipitation types (Table 4), it is found that in 215 regional extreme precipitation days, there are 72 cases of southwest vortex type, accounting for one-third, 20 cases of typhoon type, accounting for 9.3%, and 123 cases of cold trough shear line type, accounting for 57.3%. The average number of stations affected by the cold trough shear line type is the most (25 stations), and the other two types are 22 stations. The average precipitation intensity and the daily precipitation extreme value of the typhoon type are significantly stronger than those of the southwest vortex type and the cold trough shear line type, indicating that the typhoon system has good thermodynamic conditions. The daily precipitation extreme value is 540.0 mm (Hubei Yangxin station, caused by 199,406 typhoon Tim). Statistics show that 20 regional extreme precipitation days of typhoon type are related to 15 typhoons (five of which are persistent extreme precipitation caused by typhoons). The moving path of the typhoon is northwest before and after landfall. When the typhoon approaches the MYRV, its moving path is divided into two cases. If the moving path turns north or northeast, the extreme precipitation area is on the north side of the typhoon. If the typhoon continues to move northwest or turns southwest, the extreme precipitation area is on the southwest side of the typhoon.

On the meridional profile of the typhoon type (Figure 8(b)), two strong updrafts are located on the north and south sides of the typhoon center, corresponding to the two heavy rainfall areas of the north and south of the typhoon, respectively. The updraft on the south side is stronger and the extension height is higher. There is also a steep θse zone on the north side of the typhoon, but the frontal structure is loose, the position is northerly (Figure 9(b)), and the strong ascending motion mainly comes from the typhoon itself. The typhoon low-pressure circulation is affected by two strong water vapor transports, namely, the southerly water vapor transport associated with the southwest monsoon and the northerly water vapor transport associated with the typhoon low-pressure circulation. The two strong water vapor convergence centers are located on the north side of the typhoon low-pressure center (the convergence of southeast wind and northeast wind) and the southwest side (the convergence of southwest wind and northwest wind), especially on the southwest side, the convergence area is larger and the water vapor convergence is stronger, which is also the main heavy rainfall area of typhoon type.

On the meridional profile of the cold trough shear line (Figure 8(c)), the distribution of θse is similar to that of the southwest vortex type, including θse saddle field, dense zone, steep structure, and unstable stratification in the lower layer. The difference is that the isoline of θse is denser, the vertical ascending motion is stronger, and the distance between the frontal zone and the vertical ascending zone is closer, indicating that the cold air of this type is stronger, and the convergence of cold and warm airflow leads to stronger frontal zone and ascending motion. The water vapor of extreme precipitation mainly comes from the southwest warm and humid airflow. The water vapor convergence area is located in the south of the MYRV, corresponding to the heavy rainfall area, which is more southerly than the southwest vortex type (Figure 9(c)).

In this article, the climatic characteristics of extreme precipitation in the MYRV from 1981 to 2020 are statistically analyzed. The regional extreme precipitation days are classified objectively. The main weather systems and thermodynamic process characteristics of various regional extreme precipitation days are studied by using the classification synthesis method. The following conclusions are obtained:

1. The spatial distribution of the extreme precipitation threshold in the MYRV is uneven due to the regional topography, and the threshold in the south, west, and northwest of the region is lower, while higher in other regions, especially in the northeastern part of the region (greater than 90 mm). The temporal distribution of extreme precipitation days is uneven, mainly occurring from May to August. The spatial distribution of the average precipitation intensity and frequency of extreme precipitation days are characterized by the north-south antiphase distribution. The extreme precipitation intensity in the north-central part of the region is large and the frequency is low. The extreme precipitation intensity in the south is small, and the frequency is high. The intensity and frequency of extreme precipitation in the northwest are small and low.

2. In the past 40 years, there were 215 regional extreme precipitation days in the MYRV, mainly occurring from May to August, especially in June and July. According to the main influence systems, they are divided into three weather types: southwest vortex type, typhoon type, and cold trough shear line type.

3. Extreme precipitation is the result of a synergistic effect of the influencing systems. The southwest vortex type occurs in the deep warm and humid airflow in front of the southwest vortex and trough, and the rainfall area is located in the Yangtze River and its north. Affected by the typhoon's low-pressure circulation, the typhoon type has good thermodynamic conditions, and the rainfall is located on the north and southwest sides of the typhoon, especially on the southwest side. The cold and warm airflow convergence of the cold trough shear line type is more obvious, and the rainfall area is located in the south of the MYRV.

The analysis of this article is based on 315 national stations of MYRV. It can be seen from Figure 1 that the distribution of stations in the MYRV is uneven, and the station density is also inconsistent. These will have a great impact on the statistics of regional extreme precipitation days, which may affect the subsequent analysis of major weather patterns, because the same events may be recorded easily as a regional extreme precipitation day if they occur in areas with dense stations, but it is difficult to be recorded if it occurs in areas with sparse stations. At present, automatic meteorological observation stations with high spatial distribution have been built in the MYRV, but the data period is not long (less than 15 years). In the future, high-density automatic meteorological observation station data can be considered for research. On the other hand, the classification of this article is based on the first two modes of MV-EOF decomposition, and their total variance is only 33%. The reason for the low cumulative variance may be due to the small correlation between variables. The variables used in our study include the 500 hPa height field, 850 hPa wind field, and surface precipitation, and it is well known that the influencing factors of extreme precipitation are very complex, including upper trough, subtropical high, tropical cyclones (TCs) South Asia high, southwest vortex, low-level shear line, and other small and medium-scale weather systems. In addition, the influence of complex terrain in this region will also make the extreme precipitation process more complicated (Ye et al. 2007; Zheng et al. 2014; Zhang et al. 2015b). Although the classification based on several variables can reflect the main weather patterns of extreme precipitation, it cannot include all circulation situation field configurations (Chen et al. 2016; Zhang et al. 2018). Therefore, in operational forecasting, forecasters should deeply study the multiscale interaction between atmospheric circulations, especially the environmental conditions, structural characteristics, and evolution process of small- and medium-scale systems.

The study was supported by the National Natural Science Foundation of China (grant U2242201), the National Key R&D Program of China (2022YFC3002904), the Key projects of Hunan Meteorological Bureau (CXFZ2022-ZDZX01), and China Meteorological Administration Meteorological Forecast Operation Key Technology Development Project (YBGJXM(2017)1A-10).

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.