A Spatial Cluster Analysis of Heavy Rains in China

ATMOSPHERIC AND OCEANIC SCIENCE LETTERS, 2011, VOL.4, NO.1,36 -40A Spatial Cluster Analysis of Heavy Rains in ChinaTU Kai, YAN Zhong- Wei, and WANG Yi?'Key Laboratory of Regional Climate-Environment for East Asia, Institute of Atmospheric Physics, Chinese Academy of Sciences,Bejing 10029, China2National Meteorological Center, Chinese Meteorological Administration, Bejing 100081, ChinaReceived 16 November 2010; revised 10 December 2010; accepted 10 December 2010; published 16 January 2011Abstract Clustered heavy rains (CHRs) defined usingmatically from the 1970s to the 1990s. More studies arehierarchical cluster analysis based on daily observations needed to quantify climate change in terms of precipita-of precipitation in China during 1960- -2008 are investi-tion extremes. Most of the previous studies did not con-gated in this paper. The geographical pattern of CHRs insider the geographical extent or relationship between rainChina shows three high-frequency centers- South China,events recorded at individual stations. Heavy rains maythe Y angtze River basin, and part of North China aroundoccur at a local site, e.g, within 102 km2 due to localthe Bohai Sea. CHRs occur most frequently in South convectivity, or simultaneously over a larger area, e.g,China with a mean annual frequency of 6.8 (a total of334 within 10° km' in association with large scale weathertimes during 1960 -2008). June has the highest monthlyprocesses (Tao, .1980). Simultaneous heavy rains in afrequency (2.2 times/mnonth with a total of 108 times dur- large contiguous area could be clustered into a megaing 1960- 2008), partly in association with the Meiyu event of climate extremes. These kinds of large-scale ex-phenomenon in the Yangtze River basin. Within the pasttreme events often covering a large area involving dozens50 years, the frequency of CHRs in China has increased of national standard climate stations in China could besignificantly from 13.5 to 17.3 times per year, which is caused by various weather systems, such as typhoons,approximately 28%. In the 1990s, the frequency of CHRstropical disturbances, Meiyu fronts, and cyclones, espe-often reached 19.1 times per year. The geographical ex- cially during the summer monsoon season. On the othertent of CHR has expanded slightly by 0.5 stations, and its hand, isolated heavy rains at individual sites could beaverage daily rainfall intensity has increased by 3.7 mmindicative of small-scale weather systems. It is beneficiald'. The contribution of CHRs to total rainfall amount and to study the climatology and change of heavy rains withthe frequency of daily precipitation have increased bylarge geographical extents to improve our understanding63.1% and 22.7%, respectively, partly due to a significantof large-scale changes in climate extremes in China.decrease in light rains. In drying regions of North andA clustered heavy rain (CHR) is defined as a series ofNortheast China, the amounts of minimal CHRs have had heavy rains that occur simultaneously over a contiguousno significant trend in recent years, probably due toarea involving a number of adjacent stations. In this paper,warming in these arid regions enhancing atmospheric we will apply the hierarchical cluster analysis method toconvectivity at individual stations.analyze the geographical and temporal features of CHRsKeywords: cluster analysis, heavy rain, climate extremes,in China. Section 2 introduces the data and methods ingeographical correlation.detail. Section 3 discusses the results including CHR'sCitation: Tu, K, Z.-W. Yan, and Y. Wang, 2011: A spatialgeographical patterm, and changes in extent and precipita-cluster analysis of heavy rains in China, Atmos. Oceanic tion strength. Section 4 is a summary with further discus-Sci. Lett, 4, 3640.sion of the findings.1 Introduction2 Data and methodsHeavy rains, conventionally defined as those over 50We chose 549 out of 740 standard stations for dailymm d' in China, are one of the most ifuential weather obsevations of pecipitation during 1960- 2008, most ofevents associated with disasters such as floods, landslides,which are located uniformly across the easterm half ofand debrs flows In China, climatic trends for heavy rains, China. In westem China, CHRs rarely occur, parly due ta form of cimate extremes has been well studied (Zhaiet sparse observing satins.al, 2005; Han and Gong, 2003; Wang and Zhou, 2005),We define a CHR using the following process:sometimes with inconsistent results. For example, Zhai et(1) Single out every heavy rain event (≥50 mm d7) atal. (2005) found significant increases in precipitation in中国煤化工westem and southem China, but decreases in northermEysis repeatedly to theChina and the Sichuan basin. However, Tu et al. (2010) even[YHC N M H Gd longitude, and en-found that the frequency of extremely heavy rains (oversure that the minimum Euclidean distance between any100 mm d) in parts of northem China increased dra- two clusters is larger than 200 km untl all adjacent heavyrain events are clustered;Corresponding authbor: YAN Zhong-Wei, yzw@tea.ac.cn(3) Single out all clusters involving at least N (N=10).NO.1TU ET AL.: CLUSTER ANALYSIS OF HEAVY RAINS37stations as CHRs. In general, the representative area ofone standard station is nearly 10000 km^ in the eastemn459Nhalf of China. Hence, the scale of 10 stations is at least300 km as per the middle and the weather scale system,40°Nbut not in accordance with the mesoscale that causesheavy rain (Tao, 1980).350NThe hierarchical cluster analysis is a useful tool to dis-tinguish precipitation patterns in weather forecast systems(Marzban and Sandgathe, 2006) and discover the geo-309Ngraphical structure of climatology in climate studies (Fo-vell and Fovell, 1993; Qian and Qin, 2008).259NA major function in cluster analysis involves a calcula-tion of the distance or linkage between events. Th20NEuclidean distance is the most common distance metrieused in atmospheric sciences and is utilized in this paper.1009E1109E120°E130°EFirst, the distance between all stations is calculated. Sec-ond, each pair of the closest stations are merged into oneFigure 1 The cumulative frequency ofCHRs over China.cluster if the distance between them is less than 200 km.Third, the distance between a few sets of entries is recal-the station with the highest frequency of CHRs (129culated until all of the stations are clustered. The thresh-times). In Region I, there are 24 high frequency stationsold is set at 200 km because the minimum distance be- (6.9% of 350) with more than 70 CHRs during 1960-tween any two stations is less than 200 km in the eastern2008. The CHRs ocrring in at least one of these 24 sta-half of China. If the distance between any two clusters is tions account for 21.8% of all CHRs in China. In Regionlarger than 200 km, those clusters are regarded as spa-II, there is a center of high frequency (more than 70 CHRstially independent.during 1960 2008) located in the south of Anhui andFour regions in the east of China were divided by lati-northeastem Jiangxi, including 12 stations (12.8% of alltude: South China (<27°N, SC), the Yangtze River Basinstations in this region), in association with the Meiyu(27- -32°N, Yangtze), the Yellow-Huai Basin (32- 37°N,phenomenon. The CHRs occurring in at least one of theseYH), and the North (37- 42°N) and the Northeast (>429N) 12 stations account for 25.0% of all CHRs in this region.were considered as one region (NNE).In Region IIl, a center of high frcquency (more than 30times during 1960- 2008) is located in the southwest of3 ResultsLiaoning, including 12 stations (10.3% of all stations inthis region). The CHRs occurring in at least one of these3.1 Mean climatology12 stations account for 29.7% of all CHRs in this region.In China, 756 CHRs have been found during 1960-In summer, this high-frequency center could be associated2008, i.e., on an average, 15- -16 CHRs occurred everywith the windward topography in the northwest of theyear. When N increases, the more extended CHR events Gulf of Bohai.rapidly become less frequent. There are 236 CHRs ifN isTable 1 shows the monthly and regional statistics ofset at 15. The most extended CHR event occurred on theCHRs. In June, there are 205 CHRs, of which 108 CHRs19th of April 1996, involving 43 staions in six provinces are centered in the Yangtize River Basin associated withof southem China.the quasi-stationary Meiyu front. The seasonal variabilityFigure I shows a cumulative geographical coverage ofis the smallest in South China, where 50 or more CHRsCHR in China. There are 350 stations involved in at least per month occur during May-August. It is interesting thatone of the CHRs, acounting for 63.8% of the chosen 549the frequency of CHRs is much smaller in the Yel-stations. Three centers of high frequency are recognized:low-Huai Basin than in the north and northeast regionSouth China (), the area south of the Yangtve River (1)，(about twice). In the north and the northeast, CHRs occurand North China around the Bohai Sea (II) In Region I，most frequently in July and August with an annual countTaishan (#59478, 22°15'N, 112947'E) in Guangdong is of84.5%,Table I The monhly and regional frequency ofCHRs.Regioms (tations)Jan Fcb Mar Apr May Jun_ Jul_ Aug_ Sep_ Oct Nov_ Dece_ TotalSouth China (98)53122588中国煤化工s34Yangtze Basin (94)1242108TMYHCNMHG*78Yellow-Huai Basin(56)350North and Northcast (55)84Total (350)13105 205 166 122 63 2475638ATMOSPHERIC AND OCEANIC SCIENCE LETTERSVOL. 43.2 Decadal changelevel in the 1970s following a sharp climatic jump in theFigure 2 shows the long-term variation of CHRs inlate 1960s; in the Yangtze River basin, it was also rela-China and the four regions mentioned in section 2. On thetively low. In the 1990s, total rainfall in the Yangtze Riverwhole, CHRs show an increasing trend since 1960. Thebasin and South China increased remarkably, while that infrequency of CHRs has increased by 28% from 13.5 toNorth China changed little.17.3 times per year in a linear trend, during 1960- 2008.3.3 Geographical extent and precipitation intensitySimilar trends for the four regions are 36% (5.8 to 7.8) inSC, 67% (4.3 to 7.1) in Yangtze,- -52% (1.7 to 0.8) in YH,For each CHR, the geographical extent (or the numberand -13% (1.8 to 1.6) in NNE. However, according toof stations involved) and the average rainfall intensity areZhai et al. (2005), the total annual rainfall in China doestwo key parameters. A CHR with more stations and largernot show a significant trend. In the four regions, the totalintensity usually results in more serious disasters. Figureannual ainfan exhibits trends of about 3.0% (SC), 1.3% 3 shows the annual mean series for the two parameters(Yangtze), -4.3% (YH), and -12% (NNE). In NNE only,during 1960- -2008 where both extent and intensity showthe trends of CHRs' frequency and total annual rainfallan increasing trend. The geographical extent (i.e, theare comparative.number of stations) has increased from 13.4 stations in theThe CHRs frequency for China was at its lowest level early 1960s to 13.9 in recent years. The maximum meanin the 1970s, at about 12.5 times per year, and at its high-extent of CHRs (about 15.5 stations) occurred in 1997.est in the 1990s, at about 19.1 times per year. These de-With the exception of late 1970s and early 1980s, the se-cadal variations could be linked with variations in totalries of the extent exhibits an increasing trend throughout,rainfall amount. The correlation coefficients between an-especially during the 1990s. The annual mean precipita-nual total rainfall and CHR's frequency are 0.52 (SC),tion intensity of CHRs has increased from 86.5 mm in the0.60 (Yangtze), 0.37 (YH), and 0.58 (NNE), with a 95%early 1960s to 90.2 mm in recent years. The series of theconfidence level. According to Tu et al. (2010), the totalintensity has maintained an increasing trend since thsummer rainfall amount in North China was at its lowest1960s without large variations. Combined with the in-16 r1430 t2个g 250;20t4几10 I1960 1970 1980 1990 2000YearFigure 2 Annual CHRs, total (lef), and four regions (right) in China. Linear trends are given. In the right panel, solid line indicates South China,dashed line indicates Yangtze Basin, dot-dashed line indicates Yellow-Huai Basin, and dotted line indicates North and Northeast China.05 r005t95 |14 b789f Q1中著13s512 t中国煤化工d1Lo_YHCN MHG_Figure 3 The geographical extent (lent, described by the number of sations in each CHR) and precipitation inensty (right) ofCHRs.NO.1πU ET AL: CLUSTER ANALYSIS OF HEAVY RAINS3Screasing frequency of CHRs as mentioned in sub-section called minimum CHRs (N=1), i.e, those events without a3.2, the effect of changes in CHRs should be much en-correlation to any other events are also counted as CHRs.hanced in China since the 1990s.Hence, minimum CHRs include all heavy rains recorded3.4 The contribution of CHRs to total precipitationat individual stations, particularly those due to local con-vective actities. In North and Northeast China duringThe importance of CHRs can be estimated by theirthe last half century, heavy rains in summer have de-contribution to heavy rains and total rainfll amount. The creased along with three rafalldecreasing jumps (Zhaileft panel of Fig. 4 shows the percentage of heavy rainet al, 2005; Tu et al, 2010). As shown in the left panel ofevents included in CHRs for all heavy rain events. ItFig.5, the frequency of minimum CHRs in NC and NEshows an increasing trend since 1960. From 16.3% in the has decreased and increased, respetively during the lastearly 1960s to 20.0% in recent years, it has increased byhalf of the century. However, it shows no decreasing trend3.7%. In 1995, it reached a high of 32.3%. The variation in NC since the 1980s. The percentage of the minimumin heavy rain events' percentage is very close to that of CHRs in total heavy rain events shows a weak increasingthe frequency of CHRs with a correlation cefficient ofrend in NC and a stronger increasing trend in NE. Be-0.95. The right panel of Fig. 4 shows the percentage ofcause NC and NE are the most rapidly warming areas intotal rainfall of CHRs in total rainfall amount. It alsoChina, this implies that heavy rains caused by local con-shows a significant increase of 63.1% from 5.7% in thevective activities tend to be more frequent under a warm-early 1960s to 9.3% in recent years. It is noteworthy thating condition.the contribution of CHRs to total rainfall amount is notlarge except for a few abnormal years, e.g, 1969,4 Summary and discussion1994- 96, and 2008 (the highest increase at 20.3%). Mostof the annual values were well within the range from 5%CHRs are defined by using cluster analysis whichto 10% due to the very low frequency of CHRs.represents a typical type of large -scale climate extremes.3.5 Minimum CHRs in North and Northeast ChinaThere are three high-frequency centers of CHRs: SouthChina, south of the Yangtze River, and northerm ChinaAn extreme situation in calculating CHRs is the so-around the Bohai Sea. In South China and around the35 rq2005t960 1970 1980 1990 20001960 1970 19801990 2000YearFigure 4 The contribution of CHRs to total heavy rainfall events (eft) and total rainfall (right).5C8(NE4:. NC74C356030502S40 oloal R中_22(中国煤化工30中↑:YHCNMH G196019701980199020001960 1970 1980 1990 2000earFigure 5 Minimum CHRs in NC and NE and their contribution to total heavy rain events.46ATMOSPHERIC AND OCEANIC SCIENCE LETTERSVOL.4Bohai Sea, CHRs could be induced and intensified by a the Chinese Meteorological Adminitration Program (Grant No.windward topography. In the Yangize River Basin, CHRs GYH20906009).occur mainly under the quasi-stationary front during theMeiyu period. The frequency, average rainfall intensity,Referencesgeographical extent, and the contribution of CHRs to totalFovell, R., and M. Fovell, 1993: Climate zones of the conterminousrainfall amount show increasing trends during the last halfUnited States defined using cluster analysis, J. Climate, 6,of the century, while total rainfall amounts remain un-2103- -2135.changed in the monsoonal China region. The geographicalHan, H, and D. Gong, 2003: Extreme climate events over northermextent of the extreme weather systerms appears to beChina during the last 50 years,J. Geogr: Sci, 13, 469 479.slightly enlarged and enhanced under a warming back-Marzban, C., and S. Sandgathe, 2006: Cluster analysis for verifica-ground. However, the minimum CHRs in drying NC andtion of precipitation fields, Wea. Forecasting, 21, 824-838.NE remain at a certain level in recent years, indicatingQian, W, and A. Qin, 2008: Precipitation division and climate shftthat precipitation extremes at local scales may have beenin China from 1960 to 2000, Theor. Appl. Climatol, 93, 1-17.Tao, S, 1980: Torrential Rain in China, Science Press, Bejing,enhanced under warming in these regions.The identification and quantification of CHRs remainsTu, K, Z. Yan, and W. Dong, 2010: Climatic jumps in precipitationan unsolved problem, although the present results areand extremes in drying North China during 1954 -2006, J. Me-enlightening. Additional statistical methods could be use-teor. Soc. Japan, 88, 29- 42.ful in dealing with multi-scale phenomena such as pre-Wang, Y, and L. Zhou, 2005: Observed trends in extreme precipita-cipitation extremes. This research serves as a basis fortion events in China during 1961- 2001 and the associatedfurther studies of CHRs, one of the most influential cli-changes in large-scale circulation, Geophys, Res. Lett, 32,mate extremes in monsoonal Asia.L09707, doi:10.1029/2005GL022574.Zhai, P, X. Zhang, H. Wan, et al, 2005: Trends in total precipitationAcknowledgements. This study was supported by the Nationaland frequency of daily precipitation extremes over China, J.Basic Research Program of China (Grant No.2009CB421401) andClimate, 18, 1096- 1108.中国煤化工MYHCNMHG