Spatio-temporal Variability of Disastrous Convective Weather in China from 1961 to 2016

School of Public Policy and Management, Tsinghua University, Beijing 100084, China; Center for Crisis Management Research, Tsinghua University, Beijing 100084, China; Center for Social Risk Assessment in China, Tsinghua University, Beijing 100084, China

Abstract Based on the data of hail, gale, thunderstorm and lightning days in 2 481 stations in China from 1961 to 2016, the spatial and temporal distribution characteristics, periodicity and climate abruption characteristics of four kinds of disastrous convective weather in China were analyzed by various mathematical statistics methods. The results showed that in time, the days of four kinds of disastrous convective weather in China decreased, and the hail and thunderstorm days were characterized by "increasing firstly and then decreasing" from 1961 to 2016. The hail, gale, thunderstorm and lightning days in China had oscillation cycles of 3-5, 2-3, 1-2 and 1-4 a respectively, and the hail and thunderstorm days changed suddenly in 2002 and 1992 respectively. In space, the Qinghai-Tibet Plateau and western Sichuan were the high-value distribution areas of hail, gale and thunderstorm days. The high-value distribution areas of thunderstorm days were also distributed to the south of the Yangtze River. South China and its southwestern regions at the same latitude were the high-value distribution areas of lightning days. In terms of trend, the hail days in China showed a decreasing trend mainly in the Qinghai-Tibet Plateau. The gale days in China decreased in the east, was unchanged in the central region, and increased and decreased alternately in the west. The thunderstorm days in China increased in Tibet, North China, Chongqing, Zhejiang and northwestern Heilongjiang. The lightning days in China decreased obviously to the south of the Yangtze River. In terms of the fluctuation, the hail days fluctuated greatly in the southeast. The gale days fluctuated greatly to the east of Hu Huanyong line. The thunderstorms days in China fluctuated greatly in the northwest and slightly in the southeast. In addition to the small fluctuation in northern Xinjiang and South China, the lightning days fluctuated greatly in other regions of China.

Key words Disastrous weather, Hail-gale-thunderstorm-lightning, Spatio-temporal pattern, Characteristic of periodic oscillation, Mutation test

1 Introduction

Hail, gale, thunderstorms and lightning are common disastrous convective weather in China[1-2]. These strong convective weathers have characteristics of sudden occurrence, strong destructive power, strong locality, short duration, and causing disasters easily[3-6], and there are different correlations between them[7]. In the context of rapid urbanization and global climate change, disastrous convective weather has changed differently and has affected regional social production and life[8-10]. It has a significant impact on shipping transportation, agricultural production, building facilities and people’s life and property[11-13]. Among the existing disastrous convective weather, there are many studies on extreme precipitation[14-17]. In contrast, there are relatively few studies on the long-term climate change characteristics of hail, gale, thunderstorm and lightning, and there are also relatively few comparative studies on a variety of disastrous convective weather with concomitant relationships on the national scale. Most studies were conducted from the perspective of regions. Huang Yuxiaetal.[18]studied the variation characteristics of hail days in Gansu Province, and found that hail was mainly distributed in plateau and high mountains, but was rare in river valleys, basins and Gobi desert. Moreover, response analysis was performed from the perspective of weather dynamics. Yang Xiaolingetal.[19]analyzed the temporal and spatial distribution characteristics of hale days in the eastern part of the Hexi Corridor, and found that the number of gale days in the study area decreased significantly, which had a significant positive effect on the number of dust days. Zhang Yuetal.[20]studied the variation characteristics of thunderstorm in Guangdong Province and believed that urban heat island effect exacerbated the frequency of thunderstorm. Yang Chunmingetal.[21]analyzed the temporal and spatial distribution characteristics of lightning in Maanshan area, Anhui Province, and believed that the daily variation of lightning had two peaks, with the highest frequency at 14:00 and 21:00; the seasonal variation of lightning had one peak, and lightning happened most frequently in July. Most of the existing studies have started from a kind of disastrous convective weather, analyzed a disastrous convective weather process from the perspective of weather dynamics[22-23], analyzed the occurrence characteristics and influencing factors of a certain disastrous convective weather in a short period[24-25], or studied small and medium-scale convective weather forecast and topographic effects from the perspective of synoptic meteorology[26-27]. There is a relative lack of understanding of temporal and spatial distribution characteristics of various disastrous convective weather across China. With the rapid development of social economy, people’s living standard is constantly improving, and the losses and damages caused by disastrous convective weather are more severe[28-30]. Therefore, an in-depth understanding of temporal and spatial evolution characteristics of long-term disastrous convective weather events in China can not only fully understand the temporal and spatial distribution characteristics of convective weather, and make rational use of climatic resources brought about by convective weather processes to serve the national economic construction, but also has important practical significance and provides scientific and technological support for the risk assessment of disastrous convective weather disasters and reduction in the threat of disastrous convective weather to people’s life and property. In this paper, based on the data of hail, gale, thunderstorm and lightning in China during 1961-2016, the temporal and spatial distribution characteristics of the four kinds of disastrous convective weather were studied to deepen the understanding of temporal and spatial evolution rules of disastrous convective weather.

2 Data and methods

The data of days of the four kinds of disastrous convective weather (hail, gale, thunderstorm and lightning) at 2 481 stations in China during 1961-2016 were collected from theDataSetofConvectiveWeather(V1.0) of National Meteorological Information Center of China Meteorological Administration. The four weather phenomena were observed according to the requirements of Specifications for Surface Meteorological Observation[31]. If hail, thunderstorm and lightning met the requirements of the specifications, they were recorded as 1. If they did not meet the requirements, they were recorded as 0. The gale mentioned in this study refers to the wind with instantaneous wind speed reaching or exceeding 17 m/s. If it met the requirements, it was recorded as 1. If it did not meet the requirements, it was recorded as 0. The data set underwent strict consistency "Station-Provincial-National" three-level quality control. After the quality control, the missing rate of days of the four types of disastrous convective weather was less than 2%, and the accuracy of the data was close to 100%. The distribution of the stations is shown in Fig.1. Based on the above data, the temporal and spatial evolution rules of the four types of disastrous convective weather were analyzed by using linear trend, continuous wavelet power spectrum, Mann-Kendall mutation test, inverse distance weighted (IDW) interpolation and coefficient of variation[32]. Among them, linear trend, continuous wavelet power spectrum and Mann-Kendall mutation test were used to detect the variation trend, periodic oscillation and mutation characteristics of time series of the four kinds of disastrous convective weather. In the spatial differentiation diagnosis, inverse distance weighted interpolation was applied to the data of the stations. Linear trend and coefficient of variation were used to analyze the variation trend and fluctuation characteristics of the four kinds of disastrous convective weather in China from 1961 to 2016.

Fig.1DistributionofDEMandmeteorologicalstationsinChina

3 Results and analysis

3.1AnalysisoftrendcharacteristicsoftimeseriesSeen from the dynamic changes of time series, the average annual hail, gale, thunderstorm and lightning days in China from 1961 to 2016 showed different decreasing trends in fluctuations(Fig.2), and had piecewise variation characteristics. Among the four kinds of disastrous convective weather, the piecewise variation characteristics of the average annual hail and thunderstorm days in China were obvious, namely increasing firstly and then decreasing. The average annual hail days in China increased from 1961 to 1980 and decreased from 1981 to 2016 (Fig.2a). The average annual thunderstorm days in China increased from 1961 to 1972 and reduced from 1973 to 2016 (Fig.2c). China’s average annual hale days (Fig.2b) and lightning days (Fig.2d) showed decreasing trends in fluctuation, and the fluctuations in 1961-1985 were larger than that in 1985-2016.

3.2AnalysisofperiodicoscillationcharacteristicsSeen from the periodic oscillation characteristics, the four types of disastrous convective weather in China from 1961 to 2016 had different degrees of periodic oscillation characteristics in different periods. Fig.3 shows the periodic analysis of four types of disastrous convective weather days in China from 1961 to 2016 based on continuous wavelet power spectrum. In Fig.3, the zone above (inside) the V-shaped black line represents the effective spectral values of the four types of disastrous convective weather days, while the zone below (outside) the V-shaped black line represents the invalid spectral values of the four types of disastrous convective weather days. The zone surrounded by the thick black line in Fig.3 indicates that the oscillation time scale of the four types of disastrous convective weather days passed the red noise standard spectrum test at 0.05 significance level. According to Fig.3, the average annual hail days in China had an oscillation cycle of 3-5 a from 1965 to 1980 (Fig.3a). The average annual days in China had an oscillation cycle of 2-3 a from 1965 to 1975 (Fig.3b). China’s average annual thunderstorm days had an oscillation cycle of 1-2 a during 1970-1980 and 1985-1990 (Fig.3c). China’s average annual lightning days had an oscillation cycle of 1-4 years from 1965 to 1980 (Fig.3d). The oscillation cycles of the above four types of disastrous convective weather days passed the test at 0.05 significance level.

Fig.2Piecewisevariationtrendofaverageannualhail,hale,thunderstormandlightningdaysinChinafrom1961to2016

Note: a. Hail days; b. Gale days; c. Thunderstorm days; d. Lightning days.

Fig.3Periodicanalysisofaverageannualhail,wind,thunderstormandlightningdaysinChinafrom1961to2016basedoncontinuouswaveletpowerspectrum

3.3DiagnosticanalysisofmutationcharacteristicsSeen from the mutation characteristics of time series, there was a sudden change in the average annual days of the four types of disastrous convective weather in China from 1961 to 2016 (Fig.4). As shown in Fig.4, UF and UB are the MK statistic curves of average annual days of the four types of disastrous convective weather. In this study, ±1.96 and ±2.56 were taken as the MK test values of critical curves of their average annual days, and respectively characterized the confidence test interval of their average annual days at 0.05 and 0.1 significance levels. If the UF and UB statistic curves of average annual days of the four types of disastrous convective weather in China during 1961-2016 had an intersection point within ±1.96 and ±2.56, there was a sudden change in the average annual days of the four types of disastrous convective weather at 0.05 and 0.1 significance levels. According to Fig.4, the average annual hail days (Fig.4a) and thunderstorm days (Fig.4c) in China from 1961 to 2016 changed suddenly in 2002 and 1992, and passed the test at 0.1 and 0.05 significance levels respectively. The UF and UB statistic curves of average annual days of hale days (Fig.4b) and lightning days (Fig.4d) intersected in 1991 and 1988 respectively, but did not pass the test at 0.1 significance level, so the average annual days of hale and lightning days in China did not change suddenly from 1961 to 2016.

Note: a. Hail days; b. Gale days; c. Thunderstorm days; d. Lightning days.

Fig.4Mutationdiagnosisofaverageannualhail,hale,thunderstormandlightningdaysinChinafrom1961to2016basedonMKtest

3.4AnalysisofspatialdifferentiationpatternofclimaticstatesSeen from the spatial distribution of climatic states, the average annual days of the four types of disastrous convective weather in China from 1961 to 2016 had obvious regional differentiation characteristics (Fig.5). Among them, the high-value areas of average annual hail days were mainly distributed in the Qinghai-Tibet Plateau-western Sichuan and northwestern Xinjiang, where the average annual hail days were more than 4.5 d (Fig.5a). The secondary high-value areas of average annual hail days were mainly distributed in the central part of Inner Mongolia and its adjacent northern part of Hebei, where the average annual hail days were about 2.5-4.0 d. Compared with the above-mentioned regions, the average annual hail days in other parts of China from 1961 to 2016 were less than 1.5 d, and most of them were less than 1.0 d. Only in the central and eastern Inner Mongolia-central northeast China and southern southwest China, the average annual hail days reached 1.5-2.0 d. From 1961 to 2016, the high-value areas of average annual hale days in China were mainly distributed in eastern Xinjiang-the Qinghai-Tibet Plateau-western Sichuan, western and central Inner Mongolia, where the average annual hale days were mostly over 45 d (Fig.5b). The average annual hale days in vast southeastern regions and western Xinjiang were mostly less than 5 d. It is found that hail and gale were mainly distributed in plateaus (Fig.5a, Fig.5b). The average annual thunderstorm days in China showed obvious characteristics of high in the south and low in the north during 1961-2016, followed by the characteristics of high in the east and low in the west (Fig.5c). The regions with the average annual thunderstorm days over 45 d were mainly concentrated in northwestern Xinjiang, the Qinghai-Tibet Plateau-western Sichuan and the vast southern part of the Yangtze River. The average annual thunderstorm days in the central and eastern Inner Mongolia and its adjacent northwestern Shanxi and northern Hebei were also relatively more, mostly up to 30-40 d. The average annual thunderstorm days in the Yellow River-Huaihe River basin were mostly 15-20 d. The average annual thunderstorm days in the Yangtze River-Huaihe River basin were 20-35 d. Except for northwestern Xinjiang, the average annual thunderstorm days in the western part of northwestern regions were less than 10 d. China’s average annual lightning days showed a distinct north-south differentiation during 1961-2016 (Fig.5d). The high-value areas of average annual lightning days were distributed in South China and its southwestern region at the same latitude, mostly over 45 d. The average annual lightning days in the Yangtze River basin, Tibet, northwestern Xinjiang, central and northern Inner Mongolia, and most of Northeast China were around 15-25 d. The average annual lightning days in the northwestern regions were the least, mostly less than 10 d.

3.5Analysisofspatialdifferentiationpatternoflong-termchangingtrendsSpatial differentiation characteristics refer to the differences between different regions in terms of a certain attribute value. In this study, the spatial differentiation characteristics of the four types of disastrous convective weather days in China from 1961 to 2016 were analyzed by using changing trends and fluctuation characteristics based on the above spatial differentiation. The variation trends of the four kinds of disastrous convective weather days were calculated based on the least squares method, and then the IDW method was used to interpolate the four types of disastrous convective weather days (Fig.6). From 1961 to 2016, the variation trends of the four types of disastrous convective weather days had different spatial differentiation characteristics (Fig.6). The hail days in China showed different decreasing trends mainly in northwestern Xinjiang, the central part of the Qinghai-Tibet Plateau-western Sichuan, central Inner Mongolia and the central part of northeastern China (Fig.6a). The hail days in eastern and central western Tibet tended to increase slightly. Compared with the above regions, the hail days in other regions of China changed slightly. During 1961-2016, the hale days in China reduced obviously in the eastern regions, had no change in the central regions, and increased or decreased in the western regions (Fig.6b). The hale days in eastern China, mainly in central and eastern Inner Mongolia, northeastern China, eastern North China, and East China, mainly tended to decrease, mostly decreasing by more than 1.2 d/10 a. Although the hale days from the central and western Inner Mongolia to Guangxi and Guangdong in China showed a decentralized decreasing trend, the variation trends of most regions were relatively stable and not obvious. In the regions to the west of western Inner Mongolia-eastern Qinghai-western Sichuan-Yunnan, the hale days tended to increase and decrease alternately during 1961-2016. In the eastern and western parts of the regions, the hale days showed an obvious decreasing trend, mostly reducing by more than 4 d/10 a. In the central part of the regions, the hale days showed an increasing trend with point-like agglomeration, increasing by more than 4 d/10 a. The variation trend of thunderstorm days in China had obvious regional characteristics during 1961-2016. The thunderstorm days showed different increasing trends in Tibet, North China, Chongqing, Zhejiang, and northwestern Heilongjiang (Fig.6c). Among these regions, the hail days in China tended to increase mainly in Tibet and North China, and increased by more than 4 d/10 a in most of the regions. The thunderstorm days from eastern Qinghai to southwestern China, Jianghuai, Jiangnan and South China showed different decreasing trends. The thunderstorm days in South China, southwestern China and western Sichuan mostly decreased by more than 4 d/10 a. The variation trend of lightning days in China showed a distinct north-south differentiation and east-west differentiation pattern during 1961-2016 (Fig.6d). The lightning days in the regions to the south of the Yangtze River and in southern Sichuan showed a significant decrease trend, and the reduction was mostly more than 12 d/10 a. The lightning days in northeastern China, central and western Inner Mongolia, eastern North China, Huanghuai and Jianghuai regions also showed a weak decreasing trend. There was no significant change in the lightning days in the vast northwestern regions of China.

3.6AnalysisofspatialdifferentiationpatternofinterannualfluctuationcharacteristicsThe coefficient of variation, which is the ratio of standard deviation to mean, is one of the indicators to measure the degree of data dispersion or fluctuation. In this study, the coefficient of variation was used to characterize the fluctuation characteristics of the four types of disastrous convective weather days. The fluctuation characteristics of the four types of disastrous convective weather days in China showed different spatial differentiation features during 1961-2016 (Fig.7). From 1961 to 2016, the hail days in China fluctuated greatly in South China, northern Jiangnan, eastern Sichuan, Chongqing, Jianghan, Jianghuai, southern Huanghuai, western Inner Mongolia, and western Northwest China. These regions formed an inverted C-shaped structure in southeastern China (Fig.7a). Compared with the above regions, the hail days in the central Tibetan Plateau and central Inner Mongolia fluctuated slightly. The fluctuation characteristics of hale days in China during 1961-2016 were bounded by the Hu Huanyong line. The hale days fluctuated greatly to the east of the line and slightly to the west of the line in addition to southern Xinjiang and eastern Tibet (Fig.7b). In southeastern China where population and wealth are aggregated, the great fluctuation of hale days may have adverse effects on social and economic development. The fluctuation characteristics of thunderstorm days in China from 1961 to 2016 showed obvious spatial differentiation characteristics, that is, the thunderstorm days were more in the northwest and less in the southeast (Fig.7c). Among them, the high-value zones of thunderstorm days were scattered in Tibet, western Inner Mongolia and Northwest China. At the same time, the high-value zones of thunderstorm days were dotted on the north and south sides of the middle and lower reaches of the Yangtze River, northeastern China and eastern Inner Mongolia. Compared with the above regions, the thunderstorm days in other parts of China fluctuated slightly. Except that the lightning days in northern Xinjiang and South China fluctuated slightly, the fluctuation was great in other regions (Fig.7d). Among them, the lightning days in North China and eastern Tibet and on both sides of the middle and lower reaches of the Yangtze River fluctuated most greatly. In South China, thunderstorm weather is frequent, the frequency of lightning is high, so the lightning days fluctuated slightly.

Note: a. Hail days; b. Gale days; c. Thunderstorm days; d. Lightning days.

Fig.5Spatialdistributionpatternofaverageannualhail,hale,thunderstormandlightningdaysinChinafrom1961to2016

Note: a. Hail days; b. Gale days; c. Thunderstorm days; d. Lightning days.

Fig.6Spatialdistributionpatternofvariationtrendsofannualhail,hale,thunderstormandlightningdaysinChinafrom1961to2016

Note: a. Hail days; b. Gale days; c. Thunderstorm days; d. Lightning days.

Fig.7Spatialdistributionpatternofinterannualfluctuationcharacteristicsofannualhail,hale,thunderstormandlightningdaysinChinafrom1961to2016

4 Conclusions and outlooks

4.1DiscussionsandoutlooksIn the context of global warming, extreme precipitation has increased[14], and wind power has reduced in many places[5], which has gradually become a consensus. China is under the control of the East Asian monsoon system, and various convective weather activities are frequent. Due to the impact of rapid urbanization, it is of great significance to study the temporal and spatial variation characteristics of regional disastrous convective weather[11,13]. Strong convective weather occurs in small- and medium-scale weather systems with small spatial scales, and the horizontal range is generally from more than ten to two or three hundred kilometers or only tens of meters to more than ten kilometers sometimes[7]. Its life history is short with obvious suddenness, about one hour to more than ten hours, or several minutes to one hour[2]. Therefore, the disastrous convective weather, which happens suddenly, moves rapidly, and is extremely destructive, often causes serious economic losses and even casualties in human society[1,3]. Therefore, it has been highly concerned by all walks of life. It is worth noting that the data of the stations used in this paper was dense in the east and thin in the west, lacking uniformity. Because the four kinds of disastrous convective weather studied in this paper had obvious local characteristics, the spatial distribution of the stations affected the analysis results of this paper to some extent. The temporal and spatial evolution characteristics of the four kinds of disastrous convective weather need to be further analyzed under the support of grid data with uniform spatial distribution. Under the background of urbanization and global climate change, the climatic zoning of the four kinds of disastrous convective weather in China based on the mean characteristics can not meet the demands of social production and life. It is necessary to carry out the variation zoning of the four kinds of disastrous convective weather in China with the changing trend and fluctuation characteristics as indicators. At present, fettered by the data of disaster losses, although the temporal and spatial evolution characteristics of the four kinds of disastrous convective weather in China from 1961 to 2016 were discussed in this paper, the disaster situation and risk analysis of disaster-bearing bodies of population or GDP were not analyzed. Therefore, historical disaster data should be used for further comparative analysis in the future. Because of the correlation between hail and hale as well as thunderstorm and lightning, it is necessary to compare and analyze the temporal and spatial characteristics of disastrous convective weather in China. At the same time, it is necessary to further compare and analyze the characteristics of convective weather in other parts of the world and the impact of climate change and urbanization on disastrous convective weather on the basis of existing studies. On the one hand, the danger zone of high disaster-causing factors of convective weather in China and the world can be comprehensively diagnosed. On the other hand, it can provide data reference for the impact analysis and risk prevention of convective weather.

4.2Conclusions(i) In terms of time, the average annual hail, gale, thunderstorm and lightning days in China showed a decreasing trend during 1961-2016, and the hail and thunderstorm days showed a piecewise changing characteristic of "increasing firstly and then decreasing". The average annual hail, gale, thunderstorm and lightning days in China had oscillation cycles of 3-5, 2-3, 1-2, and 1-4 a respectively, and the oscillation cycles passed the test at 0.05 significance level. The average annual hail and thunderstorm days in China changed suddenly in 2002 and 1992, and passed the test at 0.1 and 0.05 significance levels respectively.

(ii) In terms of space, the high-value zones of average annual hail days in China during 1961-2016 were mainly distributed in the Qinghai-Tibet Plateau-western Sichuan and northwestern Xinjiang. The high-value zones of average annual hale days in China were mainly distributed in eastern Xinjiang-the Qingzang Plateau-western Sichuan, western and central Inner Mongolia. The average annual thunderstorm days in China showed obvious differentiation characteristics of high in the south and low in the north during 1961-2016, followed by the differentiation characteristics of high in the east and low in the west. The high-value zones of average annual thunderstorm days in China were mainly distributed in northwestern Xinjiang, the Qinghai-Tibet Plateau-western Sichuan, and to the south of the Yangtze River. The average annual lightning days in China showed a distinct north-south differentiation during 1961-2016. The high-value zones of average annual lightning days in China were mainly distributed in South China and its southwestern regions at the same latitude.

(iii) In terms of trends, the hail days in China tended to decrease mainly in northwestern Xinjiang, the middle of the Qinghai-Tibet Plateau-western Sichuan, central Inner Mongolia and the northeastern part of northeastern China. The hail days in eastern and central western Tibet showed an increasing trend. From 1961 to 2016, the hale days in China decreased in the east, were unchanged in the central part, and increase and decrease alternately in the west. The thunderstorm days in China increased in Tibet, North China, Chongqing, Zhejiang and northwestern Heilongjiang, and decreased from eastern Qinghai to southwestern China, Jianghuai, Jiangnan and South China. The lightning days in China showed a distinct north-south and east-west differentiation pattern during 1961-2016.

(iv) In terms of fluctuations, the fluctuation characteristics of the four types of disastrous convective weather in China showed different degrees of spatial differentiation during 1961-2016. These regions where the hail days fluctuated greatly formed an inverted C-shaped structure in southeastern China, while the hail days fluctuated slightly in the middle of the Qinghai-Tibet Plateau and central Inner Mongolia. The fluctuation characteristics of hale days in China during 1961-2016 were bounded by the Hu Huanyong line. The hale days fluctuated greatly to the east of the line and slightly to the west of the line in addition to southern Xinjiang and eastern Tibet. The fluctuation characteristics of thunderstorm days in China showed obvious spatial differentiation characteristics, that is, the thunderstorm days were more in the northwest and less in the southeast. The lightning days fluctuated slightly in northern Xinjiang and South China and greatly in other regions of China.