• Open Access

Tropical cyclone and extreme rainfall trends in East Asian summer monsoon since mid-20th century



[1] Proper interpretations of extreme rainfall trends in the Asian monsoon regions are complicated by tropical cyclones (TCs) from tropical oceans, whose rainfall trend may be different from the local monsoon (non-TC) rain. Here we show that the trends over the China summer monsoon region have been distorted by western North Pacific typhoons, which bring rainfall with decreasing frequency and increasing intensity. Typically the latter is not sufficient to overcome the former, so TC-related extreme rainfall trend is smaller than monsoon-related extreme rainfall. The net impact underestimates the increasing trend and overestimates the decreasing trend in monsoon extreme rainfall over most areas. The effect is minimal in the Meiyu rain belt region, but reaches 30% in northeastern and southern China. The largest distortions occur on offshore islands in the main TC paths that underwent significant decadal variations. In Hainan, the −3%/decade trend becomes +7%/decade if typhoon rainfall is removed. An opposite case occurs in Taiwan, where the extreme rainfall trend is hugely inflated by local increases in TC rainfall. These opposite effects emphasize the importance of considering the different mechanisms of rainfall systems in order to avoid mis-attribution of regional effects on extreme rainfall to thermodynamic consequences of global warming.

1. Introduction

[2] A general theoretical basis for linking global warming with increased rainfall intensity is that atmospheric water vapor capacity increases with temperature at the rate of approximately 7% per °K through the Clausius-Clapeyron equation. With this increase the rainfall intensity and extreme rainfall events may increase in a warming world [Trenberth et al., 2007] which is a major reason for expecting a positive trend in extreme rainfall observed at land stations. Recently, Min et al. [2011] showed a correspondence between simulated increases in precipitation intensity in models with anthropogenic forcing and observed extreme rainfall trends over the Northern Hemisphere land area [Groisman et al., 2005; Alexander et al., 2006]. However, this correspondence refers mainly to North America and Europe. Over the Asian monsoon regions, the positive trends are scattered [Wang and Zhou, 2005; Zhai et al., 2005; Goswami et al., 2006; Min et al., 2011; Ghosh et al., 2011].

[3] In the Asian monsoon region, significant amounts of extreme rainfall can be due to TCs [Ren et al., 2002; Kim et al., 2006, 2012; Ge et al., 2010; Lee et al., 2010], whose development and characteristics are subject to many dynamic and thermodynamic controls over tropical oceans before they approach the land areas [Emanuel, 2005; Knutson et al., 2010; Li et al., 2010; Lau and Zhou, 2012]. These controls include sea-surface temperature (SST), vertical wind shear, static stability, and tropopause temperature, so the intensity, frequency and amount of extreme rainfall over the Asian monsoon land areas can be affected by conditions over the tropical oceans. In this study we use station rainfall reports over mainland China and Taiwan to assess the trend of TC rainfall and its impact on the extreme rainfall trend during summer monsoon from 1958–2010.

2. Data and Method

[4] A total of 479 daily rainfall stations in the China mainland (including Hainan Island) and 20 stations in Taiwan are used, which cover a major part of the East Asian summer monsoon in both middle and subtropical latitudes. The stations are chosen from a total of 776 stations with the criterion that the coverage at each station is at least 95% complete over the 53 summers (June–August). These stations provide reasonably uniform and dense coverage over the China monsoon region, which includes the eastern half of China and southern provinces, and Hainan and Taiwan in the northern South China Sea (auxiliary material, Figure S1a).

[5] The 53-year average rainfall is highest along the southern coast and Taiwan, and decreases toward the north and northwest. The meridional distribution reflects the pattern of “Southern Flood and Northern Draught” (SFND) that is a prominent interdecadal phenomenon of the East Asian summer monsoon rainfall. The current phase started to develop around the 1950s [Ding, 1994]. The cause of this variation is unknown, although a variety of theories ranging from natural variability to anthropogenic forcing have been proposed [Ding et al., 2007; Zhou et al., 2009]. The extreme rainfall can be defined in two very different approaches. The methods based on extreme value theory including generalized extreme value and peak-over-threshold to infer the characterization of extremes such as 100-year return levels [Ghosh et al., 2011], or annual indexes based on the rainfall over an event, period, or threshold in the cumulative distribution function (CDF) of precipitation [Goswami et al., 2006; Min et al., 2011]. Since our purpose is to assess the impact of TCs on intense monsoon rainfall, the extreme value theory approach is not directly applicable. We therefore define extreme rainfall at each station according to the CDF of precipitation during 1958–2010 with the local thresholds determined at the 90th and 95th percentiles [Goswami et al., 2006]. The 90th percentile thresholds are shown in Figure S1b.

3. Trends of Tropical Cyclone and Monsoon Extreme Rainfall

[6] Nearly all the coastal stations are in one of the three main paths of western North Pacific TCs that move westward and influence Asia [Harr and Elsberry, 1991]. The recurved TCs move northward toward Korea and Japan, and some may affect northeastern China. The straight-moving TCs move northwestward and often follow one of two paths: a more eastern one passing over or near Taiwan and often make landfall on the southeastern coast of China, or a western one moving in the South China Sea and impacting Hainan and the southwestern coast of China. The rainfall at each station is separated into typhoon (named-TCs)-associated and monsoon (non-typhoon)-associated components. The typhoon-associated rainfall occurs not only in coastal areas and offshore islands in the paths of western North Pacific typhoons, but also in inland areas through remote connection [Harr, 2010; Galarneau et al., 2010]. These events are identified with an Objective Synoptic Analysis Technique (OSAT) [Ren et al., 2006, 2007] that uses the distance from TC center and the closeness and continuity between neighboring raining stations to trace TC-influenced rain belts that may extend from 500 km to 1100 km away from a TC center. The typhoon rainfall varies between 10% to more than 50% of the total summer rainfall in southeastern China but is also detectable over the entire China monsoon region. The percentages of (a) total rainfall and (b) extreme rainfall at each station associated with TC rainfall are shown inFigure 1. In general, typhoon rainfall contributes to a far larger percentage of the extreme rainfall than that of the total rainfall.

Figure 1.

(a) Percentage of typhoon rainfall in total rainfall and (b) percentage of typhoon rainfall in extreme rainfall.

[7] Tropical cyclone activity in the western North Pacific has been found to vary on interannual and decadal scales with SST [Chan, 2008 Knutson et al., 2010; Lau and Zhou, 2012], but the trend on the scale of half a century or more is still unclear. On the other hand, current theoretical and modeling studies mostly project that under global warming the average frequency of TCs will decrease while the cyclone intensity and rainfall will increase [Knutson et al., 2010]. The contribution to the extreme rainfall by TCs will depend on the combined effects of frequency and rainfall intensity tendencies. The 53-year linear trends of typhoon rainfall frequency and intensity using the least squares linear fit are given inFigure 2. The typhoon rainfall frequency (Figure 2a) decreases over the broad domain and reaches a maximum of 1–2 day per decade in the southern coastal region. This tendency is consistent with Lau and Zhou's [2012] report of decreasing frequency for western North Pacific TCs using satellite derived rainfall data during 1977–2007. The typhoon rainfall intensity (Figure 2b) increases over most of the southern wet part of the SFND pattern where typhoons impact more directly, with a maximum of around 5 mm day−1 per decade in Taiwan. It becomes more mixed to the north, and the trends are mainly decreasing in the northern dry part of the SFND pattern, and particularly in northeastern China. Both the observed decreasing rainfall frequency and the increasing rainfall intensity trends in the south correspond quite well with the changes of TCs over tropical oceans projected by global warming model studies [Knutson et al., 2010], although at least some of the frequency trend is likely due to the regional effect in the western North Pacific such as the relative SST [Vecchi et al., 2008; Knutson et al., 2010]. In general, the decreasing frequency more than offsetting the increasing intensity so the trend of the typhoon rainfall amount tends to decrease except between 25°N and 40°N where the signal is mixed and insignificant (not shown).

Figure 2.

The trends of typhoon rainfall (a) frequency and (b) intensity. Open circles indicate stations at or exceed 90% significance using the Student's t-test.

[8] Since typhoon rainfall contributes significantly to the summer extreme rainfall (Figure 1b), it is expected the typhoon rainfall trends will also influence the trends of extreme rainfall accordingly. To verify this, the trends of total extreme rainfall, typhoon extreme rainfall, and monsoon extreme rainfall are computed at each station. The observed total extreme rainfall trends and the monsoon rainfall trends are compared by computing their differences. By a ratio of 11:7, more stations have monsoon extreme rainfall trend larger than the total extreme rainfall trend (Figure S2). The most notable exception is in Taiwan and a few stations on the southeastern coast of the Chinese mainland, where typhoons moving on the Taiwan path likely approach or make landfall.

[9] The percentage differences between the monsoon and total extreme rainfall trends, which represent the typhoon impacts, are summarized by area-averaging over 30 5° × 5° grid boxes in the monsoon region (Figure 3) the total, typhoon, and monsoon extreme rainfall for each summer. Based on the degree of similarities in the trends among adjacent boxes and neglecting areas with small magnitudes, the data are grouped into five sectors: Hainan (HN), Southern China (SC), Yangzi Valley (YV), Central China (CC), and Northeastern China (NEC). These sectors represent the major sub-regimes of the East Asian summer monsoon rainfall, which in a normal year is characterized by the northward movement of the zonally-elongated Meiyu (Baiu in Japan and Changma in Korea) rain belt from HN and SC in early summer to YV in middle summer and to CC and NEC in late summer [Ding, 1994].

Figure 3.

The time series and linear trends of Northeastern China (NE), Yangzi River Valley (YV), Southern China (SC) and Hainan Island (HN). The different geographical sectors are defined in the top right map, which also shows the 30 5° × 5° grid boxes where trends are calculated. The table in the bottom right gives the 53-year trends of total extreme rainfall and monsoon extreme rainfall (first two columns), and the percentage reduction of the monsoon extreme rainfall due to typhoons. An infinity symbol indicates the sign of the trend is changed. Boldface indicates significance at 90% and italic-bold indicates significance at 99% using the Student's t-test.

[10] The time series of the three types of extreme rainfall along with their respective linear trends are plotted in Figure 3for all sectors except CC, which has the smallest trend (∼7%, or 1.3%/decade) over the 53 years. The table in the lower right gives the 53-year trends of total extreme rainfall and monsoon extreme rainfall (first two columns), and the percentage reduction of the monsoon extreme rainfall due to typhoons. In general the TC trend is less positive than the monsoon trend. The smallest impact occurs in YV where the positive trend of monsoon extreme rainfall is only marginally higher than that of the total extreme rainfall. In CC the monsoon trend is 11% higher, and in SC the monsoon trend is 29% higher than the total trend because the typhoon trend is near zero. The largest effect occurs in Hainan, where the total extreme rainfall decreases (−15% trend, or −2.8%/decade) but the monsoon extreme rainfall increases (35% trend, or 6.6%/decade). In NEC, well inside the northern drought part of the SFND pattern, the total extreme rainfall decreases (trend −21%, or −4%/decade) but this includes the typhoon impact that magnifies the −16% trend (−3%/decade) of the monsoon extreme rainfall by almost one third. So the typhoon impacts cause an underestimate of the positive trend, or an overestimate of the negative trend, of the monsoon extreme rainfall in nearly the entire China monsoon region. The results for the 95th percentile rainfall (not shown) are similar to the 90th percentile with the same signs and magnitudes.

[11] While only half of the trends can be considered significant, the overall underestimate of the extreme rainfall trend can be physically explained as a result of the decrease, or very small increase, of typhoon rainfall, which is itself due to the fact that the increasing trends of typhoon rainfall intensity in southern China is not sufficient to offset the overall decreasing trends of typhoon rainfall frequency. In Yangzi Valley the extreme rainfall trend for typhoon is closest to that for monsoon. Being the most active area of the Meiyu rain belt, the effect of the increasing trend of typhoon rainfall intensity may become amplified by the heavy monsoon Meiyu rainfall sufficiently to offset most of the effect of decreasing typhoon frequency.

[12] An exception to the negative impact of typhoons on the trend of extreme rainfall occurs in Taiwan (Figure 4). Because of the larger distance between Taiwan and any other stations in the China monsoon domain that renders the OSAT method less reliable, the typhoon rainfall is defined as the rainfall that occurs when an official warning by the Central Weather Bureau is in effect. Here the observed total extreme rainfall trend is near 20% over the 53 years (3.8%/decade). After removing typhoon rainfall, the monsoon component has a trend of only 11% (2%/decade), and thus there is an 80% overestimate of the extreme rainfall trend due to typhoons. For the 95th percentile extreme rainfall (Figure 4b), the observed increasing trend of 46% (8.7%/decade) totally masked the −12% (−2.3%/decade) trend for the non-typhoon rain.

Figure 4.

Same as in Figure 3 except for (a) Taiwan with the extreme rainfall threshold determined by the 90th percentile by rain events, and (b) Taiwan with the extreme rainfall threshold determined by the 95th percentile of rainfall amount. The table shows overestimate rather than underestimate.

[13] The large increasing trend of Taiwan extreme rainfall has been used as an example of super Clausius-Clapeyron scaling to suggest a rainfall trend that is drastically larger than the 7% increase per °K [Liu et al., 2009; Singleton and Toumi, 2012]. However, the very large increasing trend in Figure 4does not indicate a rainfall – global temperature relationship. It is the result of four slowly-moving heavy-rainfall typhoons during the decade 1998–2008, all of weak to medium intensity, which had taken the eastern path of a straight-moving TC. The heavy rainfall was due to the slow movements and favorable conditions for terrain interaction, and strong interactions between typhoons and a surge in the moisture-rich monsoon southwesterlies (C.-P. Chang et al., Large increasing trend of tropical cyclone rainfall in Taiwan and the roles of terrain, submitted toJournal of Climate, 2012). Since the warning is terminated as soon as a typhoon leaves Taiwan, the typhoon rainfall may be even higher than that estimated here.

4. Concluding Remarks

[14] Our results show that since middle-20th century the trend of TC extreme rainfall in most part of the East Asian summer monsoon is less than that of the monsoon extreme rainfall, except in the vicinity of the Meiyu rain belt where the two trends are similar. This is because any increase in TC rainfall intensity is insufficient to offset the decrease in TC rainfall frequency to produce a trend that matches the monsoon rainfall. Both the intensity and frequency trends are consistent with the global warming model projections. The former may be expected from the theory that precipitation intensity increases with water vapor capacity and the latter from the decrease in relative SST in the western North Pacific in the recent decades. While the present study shows that the trend of monsoon extreme rainfall is more positive than observed, the sign of the trends still varies and remains mostly negative (albeit smaller) in northeastern China after TC influences are removed. This is not surprising because of the development of the dominant multidecadal SFND pattern in the last half century. A more concise picture of the broad scale extreme rainfall trend may emerge if the regional factors that drive the SFND variation can also be isolated.

[15] The opposite drastic effects of typhoons on the trends of extreme rainfall in Hainan and Taiwan further emphasize the importance of considering the different mechanisms of heavy rainfall systems. This is necessary to avoid mingling of the effects due to regional and global drivers in the study of extreme rainfall trends [Ghosh et al., 2011].


[16] We thank Paochun Tang for help with data processing and Russ Elsberry for help to improve the manuscript. This research is sponsored by National Science Council, Taipei, Taiwan under grant NSC 100-2811-M-002-149. Additional support is provided by U. S. Naval Postgraduate School (C.-P.C.), Chinese Academy of Sciences (Y.L.), and China Meteorological Administration (X.L. and F.R.).

[17] The Editor thanks Subimal Ghosh and an anonymous reviewer for their assistance in evaluating this paper.