Weather system classification and water vapor transport characteristics of extreme value rainstorm in northwestern Hunan

Extreme rainstorm refers to the precipitation that rarely occurs in the history of a certain area. Because of its extreme nature, once it occurs, it often causes floods, mudslides, landslides, urban and rural waterlogging and other disasters, causing serious loss of life and property. In recent years, extreme precipitation has occurred and brought serious disasters around the world (Grams et al. 2014; Lynch & Schumacher 2014). In particular, under the background of global warming, global extreme rainstorm events are generally increasing (Goswami et al. 2006; Gochis et al. 2015; Takahashi et al. 2015). Some cmip5 models also show a significant increase in extreme rainstorms (Pendergrass & Hartmann 2014). In China, an extreme rainstorm is one of the most important disastrous weather events. The frequency of extreme rainstorms in most regions of China increased significantly (Zhai et al. 2005), especially in Central China, where it shows a significant upward trend due to the influence of East Asian summer monsoon and special terrain (Gao & Xie 2014). For example, the ‘7.20’ extreme rainstorm in Zhengzhou in 2021 shocked China and abroad, and many stations broke through the historical extreme value, resulting in heavy casualties and property losses (Ran et al. 2021).

Research shows that extreme rainstorm is the result of the multi-scale weather system, and occurs against the background of the typical weather situation and physical quantity factor anomaly (Kharin & Zwiers 2005; Zhou et al. 2022). The trigger mechanism and formation mechanism of an extreme rainstorm emphasized the effects of dynamic, water vapor and terrain (Zeng et al. 2020; Zhao et al. 2022). Li et al. (2019) pointed out that the influence of topography on extreme precipitation is more significant. Chen (2021) used the statistical analysis of extreme precipitation in Hunan from 1981 to 2018, according to the weather system classification: southwest vortex warm shear line type, cold trough shear line type, typhoon type, subtropical high edge type and consistent south wind type.

Continuous water vapor transport during the formation of a rainstorm is very necessary, therefore, it is more important to explore the water vapor sources and their transport channels that cause rainstorms. At present, there are many studies on water vapor transport and sources in China, but most studies use Euler's method. The water vapor flux calculated from the instantaneous atmospheric wind field also has instantaneous characteristics. The water vapor flux calculated from the instantaneous atmospheric wind field also has instantaneous characteristics, therefore, the final water vapor transport channel can only be simple, and it is impossible to quantitatively determine the contribution rate of each water vapor source to precipitation (Jiang et al. 2013). Water vapor analysis from the Lagrangian point of view performs backward tracking through the three-dimensional position of air parcels in the atmosphere, determining the three-dimensional position of the air parcel and the physical quantities such as temperature, pressure and humidity at each time can determine the trajectory and the source water vapor contribution during the water vapor transport process (Gimeno et al. 2010; Salih et al. 2015). Therefore, the analysis of water vapor from Lagrange's point of view is an indispensable and important aspect in the study of the causes of rainstorms in northwest Hunan.

The northern part of northwestern Hunan is located in the Wuling Mountains, which is northeast and southwest. The altitude is mostly above 1,000 m. The Huping Mountain (northern Shimen County) at the junction of Hunan and Hubei is as high as 2.099 m, which is the highest mountain in Hunan. At the same time, the south side of northwest Hunan is located on the northwest side of Xuefeng Mountain and the eastern edge of Yun-Gui Plateau. The complex terrain and the frequent occurrence of extreme rainstorms can easily cause geological disasters such as waterlogging, floods and debris flows, which pose a serious threat to people's lives and property safety and social and economic development. Therefore, it is of great significance to analyze the characteristics and mechanism of extreme rainstorms in northwest Hunan.

Extreme rainstorm is a small probability event, the estimation of precipitation extremes and the forecast of heavy precipitation duration are often insufficient in daily duty operations. In addition, the numerical prediction model has limited ability to predict and simulate heavy rainfall, and its mechanism is not yet fully understood. Therefore, the problem of extreme rainstorm forecasting is still a very challenging problem for operational forecasting and service work. This paper uses the existing historical site observation data and global reanalysis data, the maximum value of extreme rainstorm events in each station in northwestern Hunan was selected as the extreme value day, and the circulation situation when the rainstorm event occurred in the northwest of Hunan was analyzed. The extreme rainstorm area of each station is combined with the weather situation and elements. By analyzing the situation field of each station, the classification of extreme rainstorms in northwest Hunan is carried out and the conceptual model is summarized. Based on the analysis of physical quantities such as water vapor, combined with the backward trajectory model, the key technical indicators of extreme rainstorm forecasts in northwest Hunan are summarized. We hope to deepen the understanding of rainstorm water vapor transport in northwestern Hunan and play a role in operational applications.

The extreme value rainstorm data include the daily precipitation data of 16 national meteorological observation stations in northwest Hunan from 1961 to 2020 (Beijing time, 20:00–20:00), and ERA-5 daily reanalysis data (1° × 1°), which are taken from European Centre for Medium-Range Weather Forecasts (ECMWF). The meteorological field data Global Data Assimilation System (GDAS) data from the National Centers for Environmental Prediction (NCEP) with a resolution of 1° × 1° are used in the Hybrid Single Particle Lagrangian Integrated Trajectory model (HYSPLIT 4.9).

In the formula: P(t) is the initial position of the air parcel; P′(t + Δt) is the initial hypothetical initial position of the air parcel; P(t + Δt) is the final position of the air parcel; V(P, t) is the three-dimensional velocity vector at the initial position; V(P′, t + Δt) is the three-dimensional velocity vector of the initial hypothetical position and Δt is the time step (Draxler & Hess 1998).

When the number of calculated trajectories is huge, the pathway cannot be displayed intuitively. In addition, in order to eliminate the influence of individual abnormal trajectories, cluster analysis is used to cluster the trajectories. The method of trajectory merging and classification is based on the principle of the nearest trajectory path. Assuming that there are N trajectories, the spatial variance of each cluster is defined as the quadratic sum of the distance between each trajectory in the cluster with the corresponding point of the average trajectory of the cluster. Each trajectory is defined as an independent cluster with zero spatial variance at the starting time. The spatial variance of all possible combinations of two clusters is calculated, and any two clusters are merged into a new cluster so that the total spatial variance (TSV) of all clusters after merging is smaller than that before merging, until all trajectories are merged into a cluster. The clusters separated before the TSV increases rapidly are the final clusters, and the average trajectories of these clusters can be calculated.

The simulation area in this study is northwestern Hunan (27.5°N–30°N, 109°E–111.5°E). The initial altitudes of the simulations are six levels at 100, 500, 1,500, 2,000, 3,000 and 5,000 m. These levels roughly correspond to the near-surface layer and levels at 925, 850, 700, 800 and 500 hPa, respectively. The three-dimensional trajectories of backward tracking for 10 days are simulated four times a day (00:00, 06:00, 12:00 and 18:00), and output the hourly trajectory position (latitude, longitude and altitude) and meteorological conditions (temperature, specific humidity and geopotential height). Although the initial altitudes of the simulation are fixed, each trajectory is simulated in Lagrangian space, its altitude evolves with the circulation, so it varies with location and time.

In the formula: q represents the specific humidity; V represents the horizontal wind; ω represents the vertical velocity. ∂q/∂t represents the local variation of water vapor; ∇·qV represents the divergence of water vapor flux; ∂ωq/∂p represents the vertical transportation of water vapor; m represents the condensation rate and E represents the evaporation rate. The average water vapor budget in the region can be diagnosed by integrating the above equation from the atmospheric bottom P_s to the atmospheric top P_t and averaging the selected calculation region.

From Table 1, among the 16 counties and cities in northwestern Hunan, there were 1,301 daily extreme precipitation events in the past 60 years, with an average of 22 times per year and 81 times per station. All stations had more than 50 daily extreme precipitation processes, with the least Luxi (53 times) and the most Chenxi (115 times). The spatial distribution of the average intensity of extreme rainstorms is similar to the spatial distribution of daily extreme precipitation threshold (Figure 2(b)), with the maximum value at Luxi (123.9 mm) and the minimum value at Fenghuang (90.4 mm). The average wind speed of 10 m shows a slightly increasing distribution from south to north, and the daily mean temperature of 2 m shows a relatively average distribution. Except for Huayuan and Baojing stations, the maximum daily extreme precipitation is less than 200 mm, and other stations are more than 200 mm. The maximum value is 455.5 mm in Zhangjiajie (20:00 on July 8–20:00 on 9 July 2003), and the main maximum daily extreme values are concentrated in 200–300 mm, accounting for 69%.

From the monthly variation of frequency of daily extreme precipitation (Table 2), the daily extreme precipitation process appears from March to November, but the frequency of each month is quite different, with obvious monthly variation characteristics. The frequency is highest in May to August, with the most in July, followed by June, accounting for 30.36 and 30.28%, respectively. November is the least, accounting for 0.15%. The average intensity of daily extreme precipitation also has obvious monthly variation characteristics, with the maximum in July, followed by June, and the minimum in November. Therefore, May to August is the key season for flood control in northwest Hunan. This is related to the strong warm and humid airflow in summer and more severe convective weather (Chen et al. 2019).

Using the synoptic diagnostic analysis method, the 16 selected daily extreme rainstorm processes in northwest Hunan during the recent 70 years are classified into the Southwest Vortex and Warm Shear Line Pattern (SVWSLP), the Subtropical High Edge Pattern (SHEP) and the Cold Trough and Shear Line Pattern (CTSLP). Different from Chen et al.'s (2019) classification of extreme rainstorms in Hunan, there are only three types of extreme rainstorms in northwest Hunan. There is no Consistent Southerly Pattern and Typhoon Pattern in northwest Hunan, and these two types mainly occur in southern and eastern Hunan.

From Table 3, among the three types of extreme rainstorms in northwestern Hunan, SVWSLP has the highest frequency of occurrence, accounting for 56.25% of the total number, reaching half of the total process. It is the main type of daily extreme rainstorm in northwest Hunan, which is related to the geographical location of northwest Hunan. The second is the SHEP, accounting for 25% of the total number; and finally, is CTSLP. In terms of the average daily precipitation extremes, the three types of intensity are relatively close to each other, and the strongest is CTSLP.

In order to further discuss in detail the main weather system configuration, dynamic and thermal characteristics of each type of extreme value rainstorm process from high to low levels in northwest Hunan are studied. ERA-5 reanalysis data is used for composite analysis of the weather pattern of various types of extreme rainstorm processes at each layer in northwest Hunan.

When the forecaster is on duty, such as the southwest vortex in the eastern part of the plateau with the southwest vortex eastward impact on Hunan, and with the herringbone shear line, low-level jet and other systems together, the system is stable and less dynamic. With the special terrain of northwest Hunan, forecasters need to consider the possibility of extreme rainstorms in northwest Hunan.

The heavy precipitation of SHEP mostly appears in the midsummer season. The extreme value rainstorm in northwest Hunan needs the forecaster to pay attention to the support effect of the long-distance typhoons, the low-level shear, the ground convergence line and the unstable energy and water vapor conditions in northwest Hunan.

In the forecast of a CTSLP extreme value rainstorm in northwest Hunan, it is necessary to pay attention to the influence of a deep low trough moving eastward, where the moving speed is slow, and there are low-level cold shear and weak cold air effects.

In summary, the study of weather system classification helps forecasters to quickly identify the situation and configuration of extreme precipitation, and determine the possibility of extreme precipitation. It is essential for the prediction of extreme precipitation.

Sufficient water vapor supply is a necessary condition for rainstorms. The trajectory model HYSPLIT based on Lagrangian models can accurately track the three-dimensional position of the airflow during the transport process. Thus, the situation of water vapor transport at different levels and source water vapor contribution in the process of water vapor transport can be determined. Since the water vapor is mainly concentrated below 500 hPa, the temporal and spatial evolution of water vapor in lower and middle levels are analyzed, using clustered backward trajectories to analyze the characteristics of water vapor transport under various weather conditions. Through the above analysis, it is found that there is no frontal typhoon effect in northwest Hunan to produce extreme rainstorms, but there will be some water vapor transport in the outer cloud system of the long-distance typhoons. The previous research results have not yet involved the effect of long-distance typhoons. Therefore, in the section on water vapor transport analysis, the same type is divided into the influence of long-distance typhoons and without long-distance typhoons.

For SVWSLP with the influence of long-distance typhoons extreme value rainstorms (Figure 7(c)), there are three main water vapor transport channels in the middle and low levels. The most important channel is from the Bay of Bengal (Channel 2), which accounts for 48.83% and the contribution rates of specific humidity and water vapor flux are about 44% (Table 4), and water vapor enters northwest Hunan through the Indian Peninsula. This is followed by the Somali jet channel (Channel 6) with a faster moving speed, which accounts for 27.21%, and the contribution rate of specific humidity and water vapor flux about 22%. Finally, the South China Sea channel (Channel 1) with a slow-moving speed accounts for 23.96% and the contribution rate of specific humidity and water vapor is about 34% (Table 4). The height evolution of the water vapor channel (Figure 7(d)) shows that the initial trajectory height of all channels is low, and is uplifted to varying degrees by the influence of terrain after landing.

For SHEP with the influence of long-distance typhoons extreme value rainstorms, there are four main water vapor transport channels in the middle and low levels (Figure 8(c)). The main channel is the Somali jet channel (Channel 6) with a faster moving speed, accounting for 46.71% and the contribution rates of specific humidity and water vapor flux are about 35% (Table 4), while water vapor enters northwest Hunan from the Indian Peninsula. This is followed by the South China Sea channel (Channel 1), which accounts for 26.46% and the contribution rate of specific humidity and water vapor flux about 39%. The third is the Bay of Bengal (Channel 2), accounting for 16.05%, and the contribution rate of specific humidity and water vapor flux are about 13%. The final Western Pacific–South China Sea channel (Channel 4), accounting for 10.77%, and the contribution rate of specific humidity and water vapor are about 13% (Table 4). The height evolution of water vapor channel (Figure 8(d)) shows that all channels are uplifted to varying degrees by the influence of terrain after landing.

In summary, it can be found that the water vapor transport channels of SVWSLP all come from the low-latitude ocean surface regardless of whether it is affected by long-distance typhoons. The main water vapor transport channels of SHEP both come from the long-distance ocean surface, and are finally transported to northwestern Hunan around 650 hPa in the form of warm and wet airflow, whether affected by long-distance typhoons or not. The CTSLP is mainly characterized by a significant water vapor confrontation between the north and the south, a large temperature difference, and the baroclinicity of the atmosphere in the rainstorm area. From the height evolution of the water vapor channel, the water vapor from the South China Sea is transported to the lower atmosphere below 850 hPa, while the water vapor from the Arabian Sea is transported to the vicinity of 650 hPa. The θ_se of water vapor from the Arabian Sea and the Bay of Bengal is higher, while the θ_se of water vapor from the South China Sea and the high inland is lower, but for SHEP, the θ_se of water vapor from the high inland is larger than that of CTSLP. The analysis of the characteristics of water vapor transport and its sources under various weather types in northwestern Hunan is helpful to deepen the understanding of the characteristics of water vapor transport in northwestern Hunan and provide reference for the prediction of extreme rainstorms in northwest Hunan.

In order to quantitatively analyze the water vapor transport and budget of three different types of extreme value rainstorms in northwestern Hunan, the boundary water vapor flux and regional water vapor budget of three types of extreme value rainstorms in northwestern Hunan were analyzed.

In the past 60 years, the average daily extreme precipitation threshold in northwest Hunan is 71.6 mm, with an average of 22 times per year. The interannual variation of the frequency of extreme precipitation is similar to the interannual variation trend of the threshold, with an obvious mutation in 1995. Before 1995, both the threshold and the frequency of extreme precipitation were small, while after 1995, both the threshold and the frequency of extreme precipitation were large. The main daily extreme value rainstorm in northwestern Hunan is concentrated in 200–300 mm, and the maximum is 455.5 mm at Zhangjiajie station, which is also the daily value extreme rainstorm in Hunan Province. Extreme value rainstorms are concentrated in May–August, most in July.

Extreme value rainstorm is the result of the interaction of various influence systems, and in northwest Hunan can be divided into three categories: SVWSLP has the highest frequency of occurrence, and the typical characteristics are southwest vortex and herringbone shear in the east of the plateau at 850 hPa. For SHEP, the WPSH is strong, and northwest Hunan is on the edge of the subtropical high with a low-level shear line. For CTSLP, the trough is deep, and there is a combined effect of cold shear and weak cold air in northwestern Hunan. In work, forecasters should analyze the conditions, structural characteristics and evolution process of micro-scale systems, based on large-scale weather classification. At the same time, with reference to the numerical forecast (especially the extreme forecast of a few members of the ensemble forecast) and the comprehensive observation data, the possible falling area, intensity and starting and ending time of extreme precipitation are accurately forecasted, and the early warning signal is issued in time in the approaching stage.

The water vapor transport channels of SVWSLP with long-distance typhoons and without long-distance typhoons all come from the low-latitude ocean surface. The most important water vapor transport channel of SHEP with long-distance typhoons is from the Arabian Sea and of SHEP without long-distance typhoons is from the Bay of Bengal, which come from the long-distance ocean surface, and are finally transported to northwestern Hunan around 650 hPa in the form of warm and wet airflow. The CTSLP is mainly characterized by a significant water vapor confrontation between the north and the south, a large temperature difference, and the baroclinicity of the atmosphere in the rainstorm area. The θ_se of water vapor from the Arabian Sea and the Bay of Bengal is higher, while the θ_se of water vapor from the South China Sea and the high inland is lower, but for SHEP, the θ_se of water vapor from the high inland is larger than that of CTSLP.

For the three weather patterns, the southern boundary is the most important water vapor input boundary in the extreme value rainstorm area of northwestern Hunan, the second is the western boundary. The water vapor income in the region is positive, which means that the extreme value rainstorm area is the water vapor convergence area. A large amount of water vapor in the lower layer converges and is transported to the middle and upper troposphere through a strong vertical ascending motion to accumulate and condense, resulting in precipitation. The larger the boundary water vapor transport, the stronger the water vapor convergence in the region, and the stronger the rainfall.

Therefore, we systematically summarized the weather pattern of extreme rainstorms and analyzed the water vapor transport of each type by using backward trajectories. This can not only provide a scientific basis for accurate prediction of such events, but also is necessary for severe convective weather forecasting and pre-warning. At the same time, it is conducive to the further development and improvement of extreme precipitation forecasting technology in northwest Hunan.

The study is supported by the Key Projects of Hunan Meteorological Bureau (CXFZ2022-ZDZX01), Hunan Meteorological Bureau Innovation and Development Project (CXFZ2022-FZZX24), the Operational Capacity Building Project of Hunan Meteorological Bureau (NLJS15), Hunan Meteorological Bureau Research Business Forecasting Project (XQKJ22C008) and Hunan Meteorological Bureau Research Business Forecasting Project (XQKJ21C010). We thank Dr Zhongxi Lin for helpful discussions. Constructive comments and suggestions from two anonymous reviewers greatly improved the manuscript.

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

The authors declare there is no conflict.