Trends in the Different Grades of Precipitation over South China during 1960-2010 and the Possible Link with Anthropogenic Aerosols

Precipitation is a key climatic factor for regional climate-environment change because of its control over the surface water balance through changes in soil moisture, evapotranspiration, infiltration and runoff, all of which is taking place against the current background of increases in greenhouse gases (IPCC, 2007). Such enhanced levels of greenhouse gases can induce increases in global precipitation and, regionally, large heterogeneity of precipitation occurs across the globe (Ramanathan et al., 2001). For example, at the continental scale of Eurasia, a decline in precipitation frequency (number of wet days) has been accompanied by an increase in the annual amount (Knapp et al., 2008). This reflects the complicated spatiotemporal distribution of precipitation in contrast to temperature change, especially in the monsoon-driven regions (Dan et al., 2012) where significant human activities take place. One such region is South China (SC), which is densely populated and has experienced rapid industrialization during recent decades, especially in the megacity region of the Pearl River Delta [PRD; (21.5^°-23.5^°N, 112.5^°-115^°E); Fig.1]. The rapid growth of the population there and the associated anthropogenic activities have led to a continuous enhancement of aerosol pollution, including their precursors (Cao et al., 2003, 2004; Li et al., 2010b; Wang et al., 2011b, 2012). Exploring the linkage between precipitation and anthropogenic aerosols in SC is the objective of the present study.

SC is located in the subtropical monsoon zones, and is a region with one of the largest levels of precipitation (Yao et al., 2011) and longest duration of flooding (April to October) in China (Li et al., 2010a). Thus, it is an area susceptible to frequent occurrence of natural disasters, especially floods and droughts (He, 1998; Jian et al., 2011). The large-scale factors affecting precipitation over SC include SST, the atmospheric wind field, geographical height, the West Pacific Subtropical High, and tropical cyclones, which have been explored by many Chinese atmospheric scientists (e.g., Liang, 1994; Xin et al., 2006; Ying et al., 2011; Jian and Qiao, 2012; Yao and Qian, 2012). However, these studies focused mainly on rainy-season precipitation (Jian et al., 2008) and most concentrated more on the interdecadal variation of precipitation related to the large-scale circulation of the East Asian summer monsoon (Wu et al., 1990; Liang and Wu, 1999; Lin, 2002; Wang et al., 2004; Yang and Lau, 2004; Tong et al., 2007). The results of these large-scale studies can explain variations in intensity and frequency of precipitation events at the continental scale; however, changes in the different grades of precipitation should also be a focus of research because of the spatiotemporal heterogeneity of precipitation and the increasing occurrence of extreme weather events associated with precipitation in SC (Li et al., 2010b; Wang et al., 2011b). The changes in regional precipitation grades and associated variables (e.g., horizontal visibility, sunshine duration and low cloud cover) caused by aerosols that are already known about demonstrate the importance of carrying out studies on this topic.

Figure 1.  Locations of the 51 weather stations in SC whose data were used in the present study. The rectangle indicates the PRD region.

For example, a linkage was found between light precipitation and air pollution (Gong et al., 2007): Firstly, anthropogenic aerosols cause the amount of solar radiation reaching the ground to decrease, which reduces the radiation available for evaporation and convection in clouds. Secondly, carbonaceous aerosols absorb solar radiation and heat the lower atmosphere, which together with decreased solar radiation at the surface, results in a stabilized lower atmosphere and suppressed convection in clouds (Koren et al., 2008; Rosenfeld et al., 2008). Both observations and simulations have demonstrated that excessive aerosol particles in the air might augment droplet numbers of clouds and reduce their size, which is sure to alter the lifetime and albedo of cloud, as well as precipitation (Ramaswamy et al., 2001; Stivens and Feingold, 2009; Qian et al., 2009, 2010).

For regional climate change of different precipitation grades, some studies in other regions of China, as well as monsoon India, have been conducted in recent years. For example, (Qian and Lin, 2005) used an extreme climate index to depict statistically the distribution of precipitation, and found that a declining trend in annual mean and extreme rainfall stretches southwestward from southern Northeast China to the upper valley of the Yangtze River, while an increasing trend is presented in Xinjiang Province and Southeast China. Meanwhile, (Goswami et al., 2006) found an increasing trend of heavy rain (greater than 100 mm d^-1) and simultaneous decrease in moderate events of 5-100 mm d^-1 over central India during the monsoon seasons of 1951-2000, which resulted in no significant trend in terms of seasonal averages. (Fu et al., 2008) classified observed daily precipitation into six grades of intensity and found a declining trend of trace-precipitation days across most of China, in which a reduced trend in effective rainfall days occurs along the middle reaches of the Yangtze River valley and Southwest China, whereas an enhanced trend was reported for Northeast China, Southeast China, the eastern Tibetan Plateau and Xinjiang Province.

Regarding the relationship between aerosols and precipitation in SC, (Cheng et al., 2005) found that increased aerosol optical depth may have contributed to the known decline in precipitation in the drought region of Southeast China, which borders the east of the domain studied in the present work. (Wang et al., 2011b) pointed out that elevated aerosol loading may have weakened precipitation to less than 25 mm d^-1 in PRD during the period 2000-06. However, the long-term trends (past 50 years) in SC, including the PRD, of different precipitation grades and their relationship with aerosols have few studies. Therefore, the aims of the study were to: (1) determine the regional trends in precipitation days according to precipitation intensity over SC; (2) investigate the 50-yr trends in horizontal visibility, sunshine duration, low cloud and Gross Domestic Product (GDP); (3) investigate their linkage with aerosol optical depth (AOD); and (4) elucidate the impact of aerosols on precipitation over SC.

The daily precipitation datasets used were obtained from the China Meteorological Administration (CMA). Those stations with more than 5% of data missing for a year were filtered out, and only data subject to strict quality control covering the period 1960-2010 were used; the result was a total of 51 stations (triangles in Fig. 1). The observed daily rainfall was categorized into five grades of intensity according to CMA standards: drizzle (0≤ P<0.1 mm d^-1); small (0.1≤ P<10 mm d^-1); medium (10≤ P<25 mm d^-1); large (25≤ P<50 mm d^-1); heavy (P≥ 50 mm d^-1). Additionally, in order to investigate the trends of light precipitation impacted by aerosols, other observations including aerosol optical depth (AOD), low cloud cover, sunshine duration and horizontal visibility, were also adopted. AOD data were retrieved from MODIS satellite data during 2000-10, while data for low cloud cover, sunshine duration and horizontal visibility for the period 1960-2010 were also taken from the CMA. Furthermore, GDP data for SC spanning the period 1978-2010 were used as an indicator of the intensity of human activity and to explore the impact of socioeconomic development on precipitation.

To quantitatively study the variation of daily precipitation over SC, the trend coefficient (r_xt) method (Shi et al., 1995; Shi and Deng, 2000) was used, which determines the correlation coefficient between elements of the time sequence n (years) and natural series as 1, 2, 3…, n: where n is the time in years, x_i stands for an element magnitude in ith year, and represents the average sample variable, where =(n+1)/2. According to this method, if r_xt is positive (or negative), it means the elements in n years have an increasing (or declining) trend. The expression can be tested with n-2 degrees of freedom in the t distribution, which can demonstrate whether it is a significant climate trend, or just random variation. In addition, the statistical methods of regression analysis, tendency fit and correlation analysis (Wei, 2007) were also used in this study.

Figure 2 shows the spatial pattern of the trend coefficients of different grades of precipitation from drizzle to heavy rain. An obvious decline in drizzle and small precipitation can be found, with most stations passing the 95% confidence level, apart from several places mainly in northwestern areas for drizzle and two points (one north of 22^°N and one in the south) for small precipitation. Five stations show a slight increasing trend for drizzle, marked with small black circles, but they are not statistically significant. For medium precipitation, a complicated mix of results can be seen, with 16 stations showing an increase, but with only one passing the 95% confidence level, and 35 stations showing a decreasing trend with four passing the 95% confidence level. In other words, there is no obvious trend for this grade of precipitation. As for large and heavy precipitation, nearly all stations present an upward trend; however, few are statistically significant, meaning the increasing trend of large to heavy precipitation in SC may be a kind of natural noise.

Figure 2.  Spatial distribution of trend coefficients for 1960-2010: (a) drizzle; (b) small; (c) medium; (d) large; and (e) heavy precipitation. Stations whose trends are significant at the 95% confidence level are circled in black.

Figure 3.  Time series of (a) drizzle, (b) small, (c) medium, (d) large and (e) heavy precipitation anomalies during 1960-2010 averaged over SC and the PRD (units: d).

Figure 4.  The seasonal trend coefficients averaged over (a) SC and (b) the PRD region.

The annual mean rainy day anomalies relative to 1960-2010 averaged over SC and PRD are presented in Fig. 3. As can be seen, the drizzle trend descends markedly over this period and the variability is larger over the PRD in contrast with SC, which may be related to high levels of air pollutants induced by the strong levels of human activity there (Zhang et al., 2008). A turning point is apparent in 1987, after which the decline accelerates, and this time period corresponds to large-scale socioeconomic development of SC centralized around the PRD region (Li et al., 2007). The rainy day variation for small precipitation is similar to that of drizzle, but the variability is spread out more than for the latter. The variability of small precipitation is greater before 1987 and then shrinks thereafter, which implies the variation in this category of rainy days during 1960-87 can be mainly attributed to natural factors, e.g., the large-scale impact of a weakened monsoon (Li et al., 2010a). The rainy day trend for medium precipitation shows a slight increase over the PRD, but a small decline for SC as a whole. Meanwhile, large precipitation demonstrates a slight rising trend both over the PRD and the whole of SC, with a jump change occurring around 1985. The reason for the changes in medium and large precipitation are complicated and need deep investigation in future work. As for rainy days of heavy precipitation, a slight increasing trend can also be seen, again both over the PRD specifically and SC as a whole, and the variability after 1990 becomes greater, demonstrating that the frequencies of extreme rainfall events are increasing and can be attributed to the frequent actions of mesoscale convective systems (Wang et al., 2011a) and strong northeasterly winds from the South China Sea (Salahuddin and Curtis, 2011). The rapid urbanization that has taken place in SC also enhances low-level atmospheric heating, which is favorable for convection (Meng et al., 2007) and partly reflects the impact of strong human activities in the PRD region on local heavy rainfall.

Summary statistics of rainy days for the different grades of precipitation are presented for SC and the PRD region in Tables 1 and 2, respectively. The mean number of rainy days is largest for drizzle and small precipitation, with values of 53.04 and 107.8 for SC and 51 and 98.67 for the PRD, respectively; and the number of drizzle days in summer are 12.59 for SC and 13.01 for the PRD. The lowest mean number of days belongs to the heavy precipitation category, at about nine days for the PRD and seven days for SC. Consequently, the largest standard deviations are for drizzle and small precipitation, and the values are larger for the PRD region than for SC as a whole. The reason for this could be that air pollution involving elevated concentrations of aerosols in the atmosphere suppresses precipitation less than 25 mm d^-1 (Wang et al., 2011b), which is why the trend coefficients are negative for drizzle and small precipitation with marked statistical significance. These averages reflect the fact that an obvious declining trend for drizzle and small precipitation occurred in the PRD region during the period 1960-2010.

To investigate the seasonal changes for different grades of rainy days, Fig. 4 presents the statistically tested trend coefficients for SC and the PRD from spring to winter. Once again, a declining trend can be seen for drizzle and small precipitation across the four seasons, passing the 95% confidence level. For drizzle over SC, the largest decrease occurs in summer and the smallest decrease emerges in winter, while for small precipitation the largest decline is in autumn and the lowest in spring (the only season for which the trend was found not to be statistically significant). However, in the PRD area, the maximal decline turns out to be in summer for both drizzle and small precipitation. This decrease across all seasons reveals a potentially serious problem for the densely-populated PRD region, insofar as the decrease in drizzle and small precipitation will have a detrimental effect on local water resources, being unable to alleviate the impacts of drought, especially from autumn to early summer (Jian et al., 2008; Zhu et al., 2011).

The relatively larger grades of precipitation (medium/ large) mainly show an increasing trend, with the exception of autumn in the PRD region, and for SC the principal decreasing trend can be seen for medium precipitation. As for heavy precipitation, there is an increase across all seasons for SC, while for the PRD the largest increase is in summer. However, the changes in rainy days for the medium to heavy grades of precipitation are not statistically significant, and thus one cannot conclude an obvious trend has taken place over the past 50 years.

Naturally, the trend of precipitation is mainly affected by atmospheric structure and water vapor content, and cloud microphysical processes, such as cloud condensation nuclei (CCN), contribute greatly to the action of aerosol particles. Therefore, it is appropriate to first discuss the atmospheric circulation patterns over SC. Actually, similar studies have previously shown an increasing trend of precipitation over SC and a decreasing trend over North China, and this pattern of change has been called the "south wet/north drought" phenomenon (Gong and Ho, 2002; Hu et al., 2003).

In the present study, as shown by the spatial distributions of the trend coefficients for annual amounts of precipitation (Fig. 5a), we can see that most stations in SC show an increasing trend. The greatest increase in total precipitation can be observed over the PRD by the largest trend coefficients there, with several stations above 0.2. Meanwhile, a number of stations in the west of SC indicate a slight decreasing trend. The distribution of summer precipitation (Fig. 5b) is similar to that of annual mean precipitation because summer is the major rainy season in SC, and the highest trend coefficients are again seen over the PRD, with two stations passing the 95% confidence level. Figures 5c and d present the time series of annual and summer precipitation anomalies averaged over the SC and PRD, respectively. The increasing trend of mean precipitation averaged over all stations in SC and the PRD is very small, and the trend coefficients of annual total precipitation in SC and the PRD are 0.08 and 0.17, with 10.96 mm (10 yr)^-1 and 36.46 mm (10 yr)^-1 for their regression coefficients, respectively. Summer precipitation also shows a similar increasing trend averaged over SC and the PRD, with trend coefficients of 0.171 and 0.306, and regression coefficients of 5.12 mm (10 yr)^-1 and 14.88 mm (10 yr)^-1, respectively. The trend coefficient for summer precipitation in the PRD region passed the 95% confidence level.

Figure 5.  Spatial distribution of trend coefficients during 1960-2010 for (a) annual and (b) summer precipitation amounts, and time series of (c) annual and (d) summer precipitation anomalies during 1960-2010 averaged over SC. Stations whose trends are significant at the 95% confidence level are circled in black.

Figure 6.  MODIS satellite AOD data: (a) annual means; (b) spring; (c) summer; (d) autumn; (e) winter (units: dimensionless).

From the above analysis, it is possible that total precipitation over SC has experienced a slight increasing trend under the large-scale background of global warming and atmospheric circulation change in the past 51 years. However, the question arises: why has there been an obvious decline in drizzle and small precipitation over SC, including the PRD region? The reason is likely to relate to the impact of aerosol particles on cloud microphysical processes, which we now examine based on further data analysis.

Figure 6 presents the 2000-10 annual mean and seasonal averages of satellite AOD data from the Moderate Resolution Imaging Spectroradiometer (MODIS). It shows high AOD over land in SC, with three obvious maximum centers with values above 0.5, one of which is located in the PRD region. The maxima are much stronger in spring (Fig. 6b) and summer (Fig. 6c), and their ranges expand to a much wider extent. The largest value occurs in the PRD region, which can be inferred as being related to the strong levels of human activity there. We suggest that this spatial pattern of AOD accounts for the aforementioned decrease in drizzle and small precipitation, and the mechanism would be similar to that reported by (Stjern et al., 2011), who found an obvious signal in the effect of aerosols on light rain, defined as <1 mm d^-1, but no signal for heavy precipitation (>20 mm d^-1), during the period 1980-2008 in Europe.

However, the decrease in drizzle and small precipitation is larger in autumn than spring (Fig. 4), while AOD in autumn (Fig. 6d) is smaller compared with spring (Fig. 6b). This reflects the nonlinear relationship between AOD and light rain, and the mechanism is very similar to that found through the work of (Tan et al., 2009). They indicated that a higher AOD in spring over the PRD is not consistent with the low visibility of fog-haze weather in autumn. The discrepancy can be explained by the fact that AOD represents extinction of solar radiation by the total vertical column of aerosol, but low visibility events mainly result from extinction by near-surface aerosols within the atmospheric boundary layer. In autumn, most hydrocarbon anthropogenic aerosols have higher mixing ratios than in spring over SC, especially methane, ethane, propane and benzene, mainly from incomplete combustion of fossil fuel, biomass/biofuel including agricultural residues and coal, vehicle exhausts, natural gas, liquefied petroleum gas and biomass burning (Tang et al., 2007). Drizzle and light rain need low CCN concentrations to form (Wood, 2006; Kubar et al., 2009) in low warm cloud (Masunaga et al., 2002; Wood, 2005; Lebsock and L'Ecuyer, 2011), and thus they will be severely suppressed by the higher levels of anthropogenic air pollutants.

To further investigate the impact of meteorological conditions on drizzle and small precipitation, 51 years of data for horizontal visibility and sunshine duration were used as a substitute for the short-term availability of AOD data. The annual mean trend coefficients of observed horizontal visibility and sunshine duration are shown in Figs. 7a and b. The visibility decreases across the whole of SC and passes the significance test. Especially in the PRD region, the trend coefficient of visibility can be less than -0.8, and the reason is again likely related to the increased emissions of aerosols in SC and the PRD under the backdrop of rapid economic development, which can lead to grey-haze weather and cause a reduction in visibility. The sunshine duration shows a similar significant decline in SC and the PRD over the past 50 years due to the impact of aerosols.

While we can infer that the increased emissions of aerosols in SC over the past 50 years corresponds to the growth in anthropogenic activities during the same period, how might these rising levels of aerosols then had an impact upon precipitation? Actually, we know that aerosols can decrease the amount of solar radiation that reaches the land surface, and therefore cause less heat to be available for evaporation and energizing convective rain clouds (Ramaswamy et al., 2001). The fraction of radiation that is not reflected back to space by the aerosols is absorbed into the atmosphere, leading to heating of the air above the surface. This stabilizes the lower atmosphere and suppresses the generation of convective clouds (Koren et al., 2008). The warmer and drier air thus produces circulation systems that redistribute the remaining precipitation (Wang et al., 2004, 2011b). The drizzle and small precipitation over SC is mainly caused by convective clouds (Rosenfeld et al., 2008), and it is clear from Fig. 7c, which shows the annual mean trend coefficient of observed low cloud cover, that most stations in SC exhibit an increasing trend of low cloud cover (although some are not statistically significant), except for a small number in the northwest of the region. The enhanced low cloud indicates a higher concentration of CCN with smaller radii, leading to less cloud droplets being coalesced into drizzle and small rain, which in turn prolongs the existence of the low cloud through positive feedback. This suppressed precipitation by increased low cloud as a secondary indirect aerosol effect over SC has been suggested previously by (Cheng et al., 2005), who showed that increased aerosols may increase cloud droplet numbers and decrease the effective radius. It has also been very clearly identified in the area south of the Yangtze River in a study by (Duan and Liu, 2011). Therefore, taking all of the above analysis and observations together, the implication is that anthropogenic aerosols are a principal factor in reducing rainy days of drizzle and small precipitation in this region. The mechanism is similar to that revealed by (Wang et al., 2011b) in their 2009 simulation study using the Weather Research and Forecasting (WRF) model, which captured elevated aerosol levels leading to more cloud droplets with smaller effective radii and ultimately a decrease in light and moderate rain (<25 mm d^-1).

The interannual changes in drizzle and small precipitation also exhibit the declining trend in response to the decrease in visibility and sunshine duration (Fig. 8a). As mentioned, both horizontal visibility and sunshine duration show an obvious decline during the past 50 years, similar to the conclusion of (Wu et al., 2007), and the reason can be attributed to excessive emissions of contaminants in SC. In addition, low cloud cover averaged over SC shows an obvious increasing trend during the past 46 years (Fig. 8b), which further supports the above conclusion (Cheng et al., 2005). Finally, the GDPs of three provinces in South China (Fig. 8c) show an increase since 1985, and the value for Guangdong Province in particular, which is representative of the PRD region, shows a sharp increase after 2005. These GDP data reflect the intensity of socioeconomic development in the PRD region, adding further weight behind the notion that anthropogenic aerosols are the principal factor involved in reducing days of drizzle and small precipitation.

Figure 7.  Spatial distribution of annual trend coefficients during 1960-2010 for (a) horizontal visibility; (b) sunshine duration; and (c) low cloud cover over South China. Stations whose trends are significant at the 95%, confidence level are circled in black.

Figure 8.  Annual change in (a) sunshine duration (units: h d^-1), horizontal visibility (units: km), and drizzle and small precipitation (units: days); (b) low cloud cover (units: %); and (c) GDP of three provinces in SC (units: 0.1 billion RMB).

The aim of the present study was to investigate the linkage between trends in drizzle and small precipitation and anthropogenic aerosols in SC, including the PRD region. The indirect effect of aerosols in southern China has been proposed in previous studies (Cheng et al., 2005; Duan and Liu, 2011). We analyzed the changes in rainy days of different grades of precipitation for the period 1960-2010 and discovered an obvious declining trend for drizzle and small precipitation over SC and the PRD region. Lower levels of precipitation (<10 mm d^-1) have significantly (passing the 95% confidence level) decreased during the past 50 years, and this phenomenon appears to be closely related to increasing levels of air pollution in the form of anthropogenic aerosols, especially in summer in the PRD region when secondary organic aerosols are significantly high (Ding et al., 2012). This represents a serious environmental problem for the region that requires careful consideration. If it persists without any attempts of amelioration, it is likely to adversely affect the socioeconomic development of the PRD region and be harmful to the sustainable development of China.

In contrast to low-grade precipitation, larger precipitation levels (medium to heavy) have increased slightly over the past 50 years in the region. However, the trends did not pass the 95% confidence level, and were therefore not statistically significant. Nevertheless, the intensity of heavy precipitation over SC has been shown to have enhanced (Min and Qian, 2008), suggesting more frequent extreme events over the PRD region and SC as a whole can be expected. More research is therefore still required (particularly modeling studies) in order to explore further the mechanisms underpinning the effects of anthropogenic aerosols on changes in different grades of precipitation over SC.