Changes of summer precipitation in China: The ... - Wiley Online Library

PUBLICATIONS Journal of Geophysical Research: Atmospheres RESEARCH ARTICLE 10.1002/2014JD022456 Key Points: • We hope to know if frequency or intensity dominates the rainfall change • A new tool is used to estimate the dominance of frequency and intensity • Relative importance of moisture and air temperature is also analyzed

Changes of summer precipitation in China: The dominance of frequency and intensity and linkage with changes in moisture and air temperature Er Lu1,2, Yingting Zeng1, Yali Luo3, Ying Ding1, Wei Zhao1, Siyuan Liu1, Liqing Gong1, Ying Jiang1, Zhihong Jiang1, and Haishan Chen1 1

Correspondence to: E. Lu, [email protected]

Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science and Technology, Nanjing, Jiangsu, China, 2NOAA Climate Prediction Center, College Park, Maryland, USA, 3State Key Laboratory of Severe Weather, CAMS/CMA, Beijing, China

Abstract Citation: Lu, E., Y. Zeng, Y. Luo, Y. Ding, W. Zhao, S. Liu, L. Gong, Y. Jiang, Z. Jiang, and H. Chen (2014), Changes of summer precipitation in China: The dominance of frequency and intensity and linkage with changes in moisture and air temperature, J. Geophys. Res. Atmos., 119, 12,575–12,587, doi:10.1002/2014JD022456. Received 16 AUG 2014 Accepted 31 OCT 2014 Accepted article online 4 NOV 2014 Published online 25 NOV 2014

Using observed precipitation and the National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis, the changes in the metrics of the summer precipitation in China, the dominance of frequency and intensity of daily extreme precipitation, and the linkage with changes in moisture and air temperature are explored. Results show that over the recent 50 years, the total summer rainfall increased over the southeast and the west and decreased over the northeast. The changes in the frequency, identified with the 95% threshold and Poisson regression, and rainfall extremes show similar spatial patterns. The relative importance of the changes in frequency and intensity in the variability and changes in extreme precipitation are estimated. It is shown that, while the interannual variability of the rainfall amount is dominated by the frequency change in almost all stations, the long-term change of rainfall amount can be dominated by both frequency and intensity, depending on the station. The change in the rainfall total is linked to changes in atmospheric moisture and temperature. The results show that the variability and change of the rainfall total can be dominated by changes in both moisture and air temperature, and the relative importance depends on the region.

1. Introduction The variability and changes of precipitation have been extensively examined in previous studies through analyzing various metrics of precipitation for different global regions. For daily precipitation analysis, the metrics include the seasonal rainfall total, the precipitation frequency, and the averaged precipitation intensity [e.g., Suppiah and Hennessy, 1998; Karl and Knight, 1998; Wang et al., 2000; Gershunov and Cayan, 2003; Zhai et al., 2005; Gutowski et al., 2007; Deng et al., 2007; Wang and Zhai, 2008; Li et al., 2008; Zhou et al., 2008; Kiran et al., 2009; Teixeira and Satyamurty, 2011; Timm et al., 2011; Chou et al., 2012; Groisman et al., 2012; Chen et al., 2012; Biasutti and Yuter, 2013]. With daily precipitation, rainfall extremes have also been studied [e.g., Liebmann et al., 2001; Tank and Können, 2003; Gershunov and Cayan, 2003; Zhai et al., 2005; Wang and Zhou, 2005; Nandintsetseg et al., 2007; Zolina et al., 2009; Rodrigo, 2010; Liu et al., 2011; Piccarreta et al., 2013]. In addition, for multiday rainfalls, wet and dry spells have been examined [e.g., Arruda and Pinto, 1980; Peña and Douglas, 2002; Schmidli and Frei, 2005; Singh and Ranade, 2010; Mo and Berbery, 2011; Reiser and Kutiel, 2012; Ratan and Venugopal, 2013; Zolina et al., 2013]. In this study, we use the recent 50 year observed precipitation in 553 stations in China to first analyze the changes in the precipitation metrics, including the summer (June-July-August) rainfall total and the frequency, intensity, and rainfall amount of daily extreme precipitation. The Poisson regression [e.g., Yu and Ding, 2006; Lv et al., 2011; Villarini et al., 2013] is used to investigate the change of frequency, whose data are nonnegative and nonnormal in distribution [e.g., Cheng et al., 2003], and the result is compared with the linear trend. While assessing these fundamental changes, we will further explore the relation among the changes of the metrics. During the year-to-year variation and the long-term change in extreme precipitation, the frequency and intensity also vary, and the variability and change of rainfall amount can be attributed to the concurrent changes of frequency and intensity. The question is whether the variability and change of the rainfall amount is dominated by the change of frequency or the change of intensity. Methods have been developed in statistics to

LU ET AL.

©2014. American Geophysical Union. All Rights Reserved.

12,575

Journal of Geophysical Research: Atmospheres

10.1002/2014JD022456

analyze the dominance and estimate the relative importance in a multivariate regression [e.g., Green et al., 1978; Pratt, 1987; Budescu, 1993; Azen and Budescu, 2003; Budescu and Azen, 2004; Azen and Budescu, 2006; Gromping, 2007; Krasikova and Lebreton, 2011; Michael and Frederick, 2011]. In the present study, we will use the method proposed by Lu et al. [2010] to compare the relative importance of frequency and intensity. These changes are all for the precipitation itself. Another focus of this study is to link the change of precipitation to the Figure 1. The trends of the summer (June-July-August (JJA)) rainfall total changes in other relevant climate during the 50 years with the station data (unit: mm/d/50 yr). quantities. The change of precipitation has been generally attributed to global warming and the change in the atmospheric circulation [e.g., Chou et al., 2012; Villarini et al., 2013; Benestad, 2013]. For the formation of precipitation, the ultimate effect of the atmospheric circulation is to change the local atmospheric water vapor and air temperature and make the air saturated [e.g., Lu and Takle, 2010a; Lu and Takle, 2010b]. Thus, in this study, the linkage between the change of precipitation and the concurrent changes in the atmospheric water vapor and temperature will be examined. The method of dominance analysis will also be used to assess whether the change of moisture or air temperature is more important in the variability and change of the rainfall total. The data used in the study are introduced in section 2. The results of the changes in the precipitation metrics and the dominance of frequency and intensity are presented in section 3. In section 4, the atmospheric processes, the linkage of the change of precipitation to the changes in moisture and air temperature, and their relative importance in the variability and change of the rainfall total are analyzed. Summary and discussion are given in section 5.

2. Data The data set of observed daily precipitation, obtained from the National Climate Center (NCC) of China (http://ncc.cma.gov.cn), over the 50 years from 1961 to 2010 is used. It contains 753 stations. After quality control, removing those stations whose sites have been changed or whose record lengths are not sufficient, we include 553 stations in the calculations. The data set of gridded (0.5°×0.5°) precipitation provided by the NCC (http://cdc.cma.gov.cn) is also used when linking to the pressure level data. The National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP-NCAR) reanalysis [Kalnay et al., 1996] used in the study starts in 1948 has 17 pressure levels and a horizontal resolution of 2.5°×2.5°. The reanalysis surface temperature, specific humidity, and air temperature for the same 50 years are used.

3. Changes in Precipitation Metrics and the Dominance of Frequency and Intensity 3.1. Changes in Summer Rainfall Total and Daily Extreme Precipitation Figure 1 shows the trend of the summer rainfall total during the recent 50 years across the country, in which the Yangtze River (south) and the Yellow River (north) are plotted. Summer rainfall total has an increasing change over the Yangtze River and Huaihe River (between the Yellow River and the Yangtze River) basins, south China, and the west but has a decreasing change over north China and the northeast. Because of the possible strengthening of the hydrological cycle in the warming climate, more and more attention has been paid to precipitation extremes, which is an important portion in the rainfall total. Here we examine the extremes from daily precipitation. One of the metrics used for describing daily extreme precipitation is the frequency, which is the number of the days, in a summer, whose daily precipitation is

LU ET AL.


12,576


10.1002/2014JD022456

Figure 2. The trends of the frequency of daily extreme precipitation, identified with the 95% threshold and Poisson regression. The increasing (decreasing) trends are marked in red (blue), and the solid (open) circles indicate that the 2 trends are significant (insignificant) at the 0.05 confidence level with the χ test.

extreme according to a certain threshold. In this study, we use the threshold of 95% based on the 50 year period of record for each station. For the trend of frequency, as mentioned in section 1, it is more appropriate to use the Poisson regression [e.g., Yu and Ding, 2006; Lv et al., 2011; Villarini et al., 2013], a generalized linear model. With the n samples (for the 50 years here), the probability of the frequency in year i being ki, which takes values of 1, 2, 3, etc., follows the Poisson distribution and can be expressed as pðμi ; ; k i Þ ¼ μi ki eμi =k i !, where μi is the expectation of ki. It is assumed that the logarithm of the μi can be fitted with a linear trend [e.g., Villarini et al., 2013]. Figure 2 shows the trend of the frequency of daily extreme precipitation. Overall, the trend of frequency is similar in spatial pattern to the trend of the summer rainfall total (Figure 1), with increasing trends over the southeast and the west while decreasing trends over the northeast. This suggests that the number of the days of extreme precipitation may have a close relation with the rainfall total. There are 199 stations, among the total of 553, where the increasing trends of frequency are significant at the 0.05 confidence level, whereas there are only 25 stations where the decreasing trends are significant. What is examined in Figure 2 is the number of the days of precipitation extremes. Another metric for describing precipitation extremes is the rainfall amount from daily precipitation extremes. Generally, the more the days of precipitation extremes, the more they contribute to the rainfall total. So the trend in the rainfall amount from the extremes shown in Figure 3 is close in spatial pattern to the trend of frequency (Figure 2), and the two trends have the same signs (the same colors in the plots) in most of the stations. The increase of the rainfall amount from extreme precipitation may result from the increases in both frequency or intensity or same contribution of the two. While presenting the trends in the rainfall amount for the stations, we further divide the trends into six categories in terms of the signs of the changes in frequency and intensity (Figure 3). In this plot, for simplicity, the trend of frequency is calculated with linear regression. Comparison shows that in most stations, the trends of frequency obtained here are the same in sign as in Figure 2, which uses the Poisson regression. In some stations, however, the trends can have a different sign. The number and percentage of the stations in each of the six categories are given in Table 1. The rainfall amount from extreme precipitation has increasing trends in 326 stations, about 60% of the total (marked red). In most of them (200 stations), the increase of rainfall amount is caused by both the increase in frequency and the increase in intensity. There are 101 stations where the increase of rainfall amount is due to the

LU ET AL.


12,577


10.1002/2014JD022456

Figure 3. The trends of the rainfall amount (Tr) of daily extreme precipitation, which are divided into six categories in terms of the signs of the trends of frequency (Tt) and the trends of intensity (Tp). The ↑ (↓) in the legend represents an increasing (decreasing) trend.

increase in frequency, while intensity shows decreasing trends. There are only 25 stations where the increase of rainfall amount is caused by the increase in intensity, coupled with a decreasing frequency. On the other hand, there are 224 stations, about 40% of the total, where rainfall amount has decreasing trends (blue). In half of them (114 stations), the decrease of rainfall amount is caused by the decreases in both frequency and intensity. In 86 stations, the decrease of rainfall amount corresponds to the decrease in frequency but increase in intensity. There are only 24 stations where the decrease of rainfall amount is due to the decrease in intensity while frequency increases. 3.2. The Dominance of Frequency and Intensity in Changes of Extreme Precipitation In a large portion of the stations, as mentioned above, the increase of rainfall amount is caused by increases in both frequency and intensity. Then, the question is whether the increase of frequency or the increase of intensity contributes more in the increase of the rainfall amount. Similarly, the interannual variability of the rainfall amount may result from both the interannual variation of frequency and the interannual variation of intensity. Then, which of the variations in frequency and intensity contributes more in the interannual variability of the rainfall amount? Table 1. The Number and Percentage of the Stations in Each of the Six Categories, With the Increasing (Decreasing) Trend in the Rainfall Amount (Tr) of Extreme Precipitation Being Caused by Increasing (Decreasing) Trends in Both Frequency (Tt) and Intensity (Tp), an Increasing (Decreasing) Frequency Coupled With a Decreasing (Increasing) Intensity, or an Increasing (Decreasing) Intensity Coupled With a Decreasing (Increasing) Frequency Catalog Tr↑Tp↑Tt↑ Tr↑Tp↓Tt↑ Tr↑Tp↑Tt↓ Tr↓Tp↓Tt↓ Tr↓Tp↑Tt↓ Tr↓Tp↓Tt↑

LU ET AL.

Number of Stations

Percentage (%)

200 101 25 114 86 24

36.4 18.4 4.6 20.7 15.6 4.4


These belong to the general issue of statistics, regarding analyzing the dominance and estimating the relative importance or contribution of predictors in a multivariate regression. Several methods have been developed in previous studies to estimate the relative importance [e.g., Green et al., 1978; Pratt, 1987; Budescu, 1993; Azen and Budescu, 2003; Gromping, 2007; Krasikova and Lebreton, 2011; Michael and Frederick, 2011]. Lu et al.

12,578


10.1002/2014JD022456

(a)

(b)

Figure 4. The (a) spatial distribution of the comparison of St and Sp (the stations with St > Sp are marked in red and otherwise in blue) and (b) scatterplot of St and Sp. The unit is 100 mm/d.

[2010] proposed a novel method to investigate the issue, which has been applied in climate studies [e.g., Hua and Chen, 2011; Li, 2012]. Here we use this method to address the above questions. The relation of the rainfall amount (r) is a function of frequency (t) and intensity (p), r = r(t, p). This relation is nonlinear, but for simplicity, a linear regression is generally used to approximate it. By using the data of these quantities for the 50 summers, the ∂r/∂t and ∂r/∂p can be calculated and used to form linear regression of the three quantities. We perform a t test for the regression equation and an F test for the regression coefficients. Results show that for the relation of rainfall amount with frequency and intensity, the regression equation and coefficients are both significant at the 95% confidence level for all the stations (figures not shown). For the interannual variability, we can use σ t and σ p, the standard deviations of t and p determined from the data, to indicate their magnitudes. The products of the change rates and the variation scales, i.e., St ≡ |∂r/∂t| σ t and Sp ≡ |∂r/∂p| σ p, can be used to measure the variabilities of r induced by the variations of t and p. LU ET AL.


12,579


10.1002/2014JD022456

(a)

(b)

Figure 5. The (a) spatial distribution of the comparison of Wt and Wp (the stations with St > Sp are marked in red and otherwise in blue) and (b) scatterplot of Wt and Wp. The unit is 0.01 mm/d/50 yr.

Figure 4a presents the result of the comparison between St and Sp. At most stations in China, except for the few stations in the west, the St is greater than Sp (Figure 4a). The scatterplot (Figure 4b) further shows that the St is much greater than the Sp. This suggests that in the interannual variability of the rainfall amount, the year-to-year variation of frequency is much more important than the year-to-year variation of intensity. Sun et al. [2007] also found that for very heavy precipitation, the percentage increase in frequency is much larger than the percentage increase in intensity. An understanding of this result is that, since the intensity of the daily extreme precipitation defined with the threshold of 95% is already sufficiently strong (relative to the specific station), there might be less space for the intensity to become stronger. However, since the frequency corresponding to the strong intensity is small, 1 day more or 1 day less in the frequency may lead to a large change in the contribution to the interannual variability of the rainfall amount. This result demonstrates the reliability of the simple method for showing the relative importance of the components.

LU ET AL.


12,580


10.1002/2014JD022456

For long-term changes, we use |Δt| and |Δp|, the absolute values of the long-term changes of t and p, for the scales of their changes. The products of the change rates and the change scales, i.e., Wt ≡ |∂r/∂t| |Δt| and Wp ≡ |∂r/∂p| |Δp|, are used to measure the scales of the changes of r induced by the changes of t and p. Unlike the result of the interannual variability (Figure 4a), there is no clear spatial pattern in the result of the long-term changes (Figure 5a). The comparison of Wt and Wp indicates that both the change of frequency and the change of intensity can dominate the long-term change of the rainfall amount and whether frequency or intensity dominates the change depends on the station. Across the country, there are more stations where the change of rainfall amount is dominated by frequency. The scatterplot (Figure 5b) shows that in those stations where the Wt is large, frequency dominates the change. At individual stations in Figure 3 where rainfall amount has an increasing (decreasing) change and the change is caused by the increasing (decreasing) changes in both frequency and intensity, the change of rainfall amount can be dominated by frequency in some of these stations and dominated by intensity in others. In some stations in Figure 3, rainfall amount has an increasing change, and this change is caused by the increase of frequency, while intensity decreases. Here the decrease of intensity has a negative effect on the increase of rainfall amount, but the increase of frequency has a stronger positive effect. In this case, the change of rainfall amount is controlled by the frequency. Comparing Figure 5a with Figure 3, we find that the dominance of frequency at these stations is captured. It also works correctly in the stations where the increase of rainfall amount is caused by increase of intensity but with a decrease in frequency, as well as the stations where the decrease of rainfall amount is caused by decrease of frequency (intensity) but with increase in intensity (frequency). These results also demonstrate the robustness of the tool we use.

4. Linking the Change of Rainfall Total to the Changes in Moisture and Air Temperature 4.1. Atmospheric Processes and Climate Quantities Relevant to the Change of Rainfall Surface warming has been stressed in previous studies probably because of the importance of surface evaporation and its impact on humans. Figure 6a shows the trend of the near-surface air temperature. Warming is strong over east China, from the south to northeast, but is relatively weak over the west. In the central region and part of the west, there is significant cooling. When linking to the change of precipitation, since saturation and condensation occur in the atmospheric levels, we should also consider the change of the air temperature. Figure 6b shows the trend of the air temperature at 700 hPa. The spatial pattern of this plot is similar to that of the surface temperature (Figure 6a). The major difference is that the cooling in the central becomes stronger and the area becomes wider. Thus, the changes in the surface temperature and the air temperature at 700 hPa are not entirely consistent. Note that this is for summer. We also plotted the trend of the annual mean air temperature, and overall, there is warming across China (figure not shown). The change of precipitation has been linked to global warming in previous studies [e.g., Wang et al., 2006; Chou et al., 2012; Benestad, 2013]. Precipitation changes are also associated with local warming [e.g., Villarini et al., 2013]. If the local warming is strong, comparative to the surroundings, the moisture over the area may increase through the enhanced thermally induced atmospheric circulation, which may bring more water vapor. On the other hand, the strong atmospheric warming relative to the surface warming may reduce the instability and the vertical transport of the water vapor. Due to the effect of the local warming and the influences of the surface warming and the surrounding air warming, the trend of the atmospheric moisture (Figure 6c) displays both similarity and dissimilarity, in spatial distribution, to the trend of air temperature (Figure 6b). For example, compared with the change of air temperature, moisture also has increasing changes over the southeast and an area over the southwest, but its decreasing change can have a larger area over the northeast and the northwest. In addition to the possible contribution to the moisture, the local air temperature specifies the saturation point of the air. The change of relative humidity (Figure 6d) is influenced by both the changes in moisture (Figure 6c) and air temperature (Figure 6b). The three trends are all positive over the southeast. This suggests that, although air temperature has increasing change, the increase of moisture is larger, and relative humidity has increased. Over the northeast, water vapor is decreasing as temperature is increasing, so both contribute to the

LU ET AL.


12,581


(a) Ts

(b) T700

(c) q700

(d) rh700

10.1002/2014JD022456

Figure 6. The trends of summer (JJA) (a) surface temperature (Ts, K/50 yr) and the trends of summer (b) air temperature (T700, K/50 yr), (c) specific humidity (q700, g/g/50 yr), and (d) relative humidity (RH700, 1%/50 yr) at 700 hPa.

decrease in relative humidity. Over the central region, there is decrease in moisture, but the decrease of temperature is larger, and thus, relative humidity increases. Lu and Takle [2010b] showed that the change of the seasonal rainfall total and the change of the seasonal mean relative humidity have a strong positive relationship. Comparing Figure 6d with Figure 1, we observe that in east China, which is relatively wet, rainfall total and relative humidity both are increasing over the region from the Yangtze River basin to south China, and both are decreasing over the region from the Yellow River basin to the northeast. The above analysis suggests that the change of precipitation, which may be related to warming, is more directly linked to the changes in atmospheric moisture and temperature. In the following section, we will examine whether the variability and change of the rainfall total is dominated by the change in moisture or the change in air temperature. 4.2. The Dominance of Moisture and Air Temperature in the Change of Rainfall Total Here we evaluate the relation of summer rainfall total (R) with specific humidity (q) and air temperature (T), R = R(q, T). Using the data of the 50 summers, the ∂R/∂q and ∂R/∂T can be obtained from the linear regression of the three quantities. The t test shows that for the relation of the summer rainfall total with moisture and air temperature, the regression equation is significant at the 90% confidence level for most of the stations, and the F test shows that the regression coefficients are significant at the 90% confidence level for about one third of the stations (figures not shown). For interannual variability, we use the products of the change rates and the standard deviations, i.e., Sq ≡ |∂R/∂q| σ q and ST ≡ |∂R/∂T| σ T, to measure, respectively, the scales of the variability of rainfall total induced by the variations of moisture and air temperature. The total contribution to the variability from moisture and air temperature is measured with SqT = (Sq2 + ST2)1/2. LU ET AL.


12,582


10.1002/2014JD022456

(a)

(b)

Figure 7. The (a) spatial distribution of SqT (the stations with Sq > ST are marked in red and otherwise in blue, and the values of zero are from the rounding) and (b) scatterplot of Sq and ST. The unit is 100 mm.

Figure 7 shows the comparison between the Sq and ST. The scatterplot suggests that the year-to-year variations of both moisture and air temperature are important to the interannual variability of the summer rainfall total and that the relative importance depends on station. The spatial distribution of the SqT in Figure 7a indicates that for the stations over south China, especially along the south coast, the variability is dominated by air temperature. In this region, water vapor can be expected to be sufficient in summer, and cold air surges can be key to the rainfall events. In the stations over the northeast, the variability of the rainfall total is also dominated by air temperature. It is hard for the summer monsoon to transport water vapor to this region, and the air there is generally dry [e.g., Lu et al., 2011; Lu et al., 2014]. The air temperature of this region, which is affected by the cold air surges, might vary more from year to year. In most stations over the central east, in the Yangtze River and Yellow River basins, the variability of rainfall is dominated by the variation of moisture. Previous studies revealed that, although a cold air surge is also

LU ET AL.


12,583


10.1002/2014JD022456

(a)

(b)

Figure 8. The (a) spatial distribution of WqT (the stations with Wq > WT are marked in red and otherwise in blue, and the values of zero are from the rounding) and (b) scatterplot of Wq and WT. The unit is 10 mm.

required, the hydrological extremes (i.e., floods and droughts) over the central region largely depend on whether the southwest monsoon and the southeast monsoon, controlled by the western Pacific subtropical high, can bring moisture to the region [e.g., Lu et al., 2014]. The results of the dominance analysis at other condensation levels (figures not shown) are similar in spatial pattern, although there may be differences for some stations. For the long-term changes, we use the products of the change rates and the |Δq| and |ΔT|, the absolute values of the changes in moisture and air temperature, i.e., Wq ≡ |∂R/∂q| |Δq| and WT ≡ |∂R/∂T| |ΔT|. These measure, respectively, the scales of the changes of rainfall totals induced by the changes in moisture and air temperature. The total contribution to the change from moisture and air temperature is measured with WqT = (Wq2 + WT2)1/2. Figure 8 presents the comparison between the Wt and Wp. The scatterplot shows that the changes of both moisture and air temperature are important to the long-term change of summer rainfall totals, and the

LU ET AL.


12,584


10.1002/2014JD022456

relative importance depends on station. The spatial distribution of the WqT indicates that in most stations south of the Yellow River, the change of rainfall total is dominated by the change of air temperature. Especially, in those stations, which are located along the southwest coast and have large values of WqT, the change of air temperature is much more important than the change of moisture (Figure 8b). Over the southeast coast, in Figure 6b and Figure 1, the air temperature and the rainfall total are both increasing. As mentioned above, the atmospheric warming there may contribute to the increase of the water vapor over the region (Figure 6c). The regional strong warming may strengthen the horizontal thermal contrast and thus the circulation, which may advect more moisture from over the ocean. However, in most stations over the northeast, the change of rainfall total is dominated by the change of moisture (Figure 8a). Over the northwest, from Figure 6c and Figure 1, the moisture and the rainfall total are both decreasing. The above results show that both the interannual variability and the long-term change of the summer rainfall totals can be dominated by both moisture and air temperature. This is different from the conclusion for the dominance of frequency and intensity, measured using the same assessment tool. For the latter, the interannual variability of the rainfall amount of extreme precipitation is dominated by frequency, while the change of the rainfall amount can be dominated by both frequency and intensity.

5. Summary and Discussion Global warming may lead to intensification of the hydrological cycle [e.g., Ziegler et al., 2003; Yang et al., 2003; Stocker and Raible, 2005; Wu et al., 2005] and thus may lead to changes in both the seasonal rainfall totals and the frequency and intensity of precipitation extremes. Because of this, numerous studies, mentioned in section 1, have been made to assess the changes in the amount, frequency, and intensity of precipitation over different regions. Using the observed precipitation from the recent 50 years, we calculated the changes of these precipitation metrics at a station scale. To be more appropriate, we use Poisson regression to determine the trend of frequency and compare it with the linear trend. Results show that the trends of the frequency and the rainfall amount of daily extreme precipitation are similar in spatial pattern to the trend of total summer rainfall, with increasing changes over the southeast and decreasing changes over the southwest. In previous studies, the changes of the precipitation metrics were mostly examined individually. The connections between the metrics, e.g., the relation of the change in rainfall amount with the change of frequency or the change of intensity, have been less studied. The variability and change of the rainfall amount can be affected by the changes in both frequency and intensity, and we may need to consider the response of the rainfall amount to the concurrent changes in frequency and intensity. A goal of this study is to understand whether the change in frequency or the change in intensity is more important to the variability and change of rainfall amount. For this purpose, a method is required to analyze which factor dominates. A simple and reasonable method, proposed by Lu et al. [2010], is applied here to estimate the relative importance of the changes in frequency and intensity. Results show that for the daily precipitation extremes defined with the threshold, the interannual variability of their rainfall is dominated by the variation of frequency in almost all stations. However, the long-term change of the rainfall amount can be dominated by both the change in frequency and the change in intensity, depending on the station. The results of the estimation of relative importance based on different assessment tools may have differences, especially when the contributions of the variables are similar. Our results, as presented in section 3.2, manifest that the tool we use in this study is reliable and robust. The above analyses of the changes and the relative importance of the precipitation metrics are all for the precipitation itself. As a related issue, the linkage of the variability and change of the rainfall total with the changes in moisture and air temperature, the most relevant climate quantity, is also explored. Physically, to form precipitation, the atmosphere needs to be saturated, and thus, the rainfall total can be influenced by both the moisture and the air temperature. As another focus of this study, we evaluate which of the two different quantities is more important. The same method for dominance analysis is applied to estimate the relative importance of the changes of moisture and air temperature in the variability and change of the rainfall total. It is shown that the regression of the summer rainfall amount with frequency and intensity is statistically significant for all stations, while the linear relation of the summer rainfall total with moisture and air

LU ET AL.


12,585


10.1002/2014JD022456

temperature is not very strong, as assessed from all the stations. The reason is that the rainfall amount is completely determined by the frequency and intensity. However, the seasonal rainfall total might not be completely determined by the seasonal mean moisture and air temperature. There is uncertainty in the relation, and the rainfall total may also be affected by other factors. What is presented in section 4 is the preliminary result of the linkage, and further investigations are required. The change of precipitation has been generally attributed to global warming, which may cause changes in the atmospheric circulation. Several atmospheric processes are analyzed in this study, to illustrate that the change of precipitation can be affected by the warming in both the surface and the upper levels, and both locally and regionally. As our next work, a physical framework will be developed to better understand the linkage between the regional change of precipitation and the warming, through considering the direct and indirect effects of the regional warming and their positive and negative contributions to the regional changes of precipitation. Acknowledgments This study was supported by the National Natural Science Foundation of China (grants 41275092, 41230422, and 41230528), National Basic Research (973) Program of China (grants 2013CB430203 and 2012CB955301), NOAA Climate Prediction Center, and the Priority Academic Program Development of Jiangsu Higher Education Institutions. The two anonymous reviewers and Ruby Leung, the Editor, are thanked for their constructive suggestions that helped improve the manuscript. The data of precipitation used in this study were provided by the National Climate Center (NCC) of China. The reanalysis data were provided by the National Centers for Environmental Prediction (NCEP) and the National Center for Atmospheric Research (NCAR).

LU ET AL.

References Arruda, H. V. D., and H. S. Pinto (1980), An alternative model for dry-spell probability analysis, Mon. Weather Rev., 108, 823–825. Azen, R., and D. V. Budescu (2003), The dominance analysis approach for comparing predictors in multiple regression, Psychol. Method, 8, 129–148. Azen, R., and D. V. Budescu (2006), Comparing predictors in multivariate regression models: An extension of dominance analysis, J. Behav. Stat., 31, 157–180. Benestad, R. E. (2013), Association between trends in daily rainfall percentiles and the global mean temperature, J. Geophys. Res. Atmos., 118, 10,802–10,810, doi:10.1002/jgrd.50814. Biasutti, M., and S. E. Yuter (2013), Observed frequency and intensity of tropical precipitation from instantaneous estimates, J. Geophys. Res. Atmos., 118, 9534–9551. Budescu, D. V. (1993), Dominance analysis: A new approach to the problem of relative importance of predictors in multiple regression, Psychol. Bull., 114, 542–551. Budescu, D. V., and R. Azen (2004), Beyond global measures of relative importance: Some insights from dominance analysis, Organ. Res. Method, 7, 341–350. Chen, C., C. Chou, and C. Chen (2012), Regional perspective on mechanisms for tropical precipitation frequency and intensity under global warming, J. Clim., 25, 8487–8501. Cheng, B., Y. Ding, and F. Wang (2003), A diagnosis method of the extreme features of weather and climate in time series based on nonnormal distribution, Chin. J. Atmos. Sci., 27, 920–928. Chou, C., C. Chen, P. Tan, and K. Chen (2012), Mechanisms for global warming impacts on precipitation frequency and intensity, J. Clim., 25, 3291–3306. Deng, Y., K. P. Bowman, and C. Jackson (2007), Differences in rain rate intensities between TRMM observations and community atmosphere model simulations, Geophys. Res. Lett., 34, L01808, doi:10.1029/2006GL027246. Gershunov, A., and D. R. Cayan (2003), Heavy daily precipitation frequency over the contiguous United States: Sources of climatic variability and seasonal predictability, J. Clim., 16, 2752–2765. Green, P. E., J. D. Carroll, and W. S. DeSarbo (1978), A new measure of predictor variable importance in multiple regression, J. Mark. Res., 15, 356–360. Groisman, P. Y., R. W. Knight, and T. R. Karl (2012), Changes in intense precipitation over the central United States, J. Hydrometeorol., 13, 47–66. Gromping, U. (2007), Estimators of relative importance in linear regression based on variance decomposition, Am. Stat., 61, 139–147. Gutowski, W. J., E. S. Takle, K. A. Kozak, J. C. Patton, R. W. Arritt, and J. H. Christensen (2007), A possible constraint on regional precipitation intensity changes under global warming, J. Hydrometeorol., 8, 1382–1396. Hua, W., and H. Chen (2011), Response of land surface processes to global warming and its possible mechanism based on CMIP3 multimodel ensembles [in Chinese], Chin. J. Atmos. Sci., 35, 121–133. Kalnay, E., et al. (1996), The NCEP/NCAR 40-Year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437–471. Karl, T. R., and R. W. Knight (1998), Secular trends of precipitation amount, frequency, and intensity in the United States, Bull. Am. Meteorol. Soc., 79, 231–241. Kiran, C., B. Krishnamurthy, U. Lall, and H. H. Kwon (2009), Changing frequency and intensity of rainfall extremes over India from 1951 to 2003, J. Clim., 22, 4737–4746. Krasikova, D., and J. M. Lebreton (2011), Estimating the relative importance of variables in multiple regression models, Int. Rev. Ind. Organ. Psychol., 26, 174–192. Li, H. (2012), The analysis of variation characteristics and cause of drought-wetness over Tarim River Basin in recent 50 years [in Chinese], PhD dissertation, pp. 115, Nanjing Univ. of Information Science and Technology. Li, H., T. Zhou, and R. Yu (2008), Analysis of July-August daily precipitation characteristics variation in eastern China during 1958–2000, Chin. J. Atmos. Sci., 32, 358–370. Liebmann, B., C. Jones, and L. M. V. Carvalho (2001), Interannual variability of daily extreme precipitation events in the State of São Paulo, Brazil, J. Clim., 14, 208–218. Liu, B., M. Henderson, M. Xu, and Y. Zhang (2011), Observed changes in precipitation on the wettest days of the year in China, 1960–2000, Int. J. Climatol., 31, 487–503. Lu, E., and E. S. Takle (2010a), Concurrent variations of water vapor and temperature corresponding to the interannual variation of precipitation in the North American Regional Reanalysis, J. Geophys. Res., 115, D11101, doi:10.1029/2009JD012956. Lu, E., and E. S. Takle (2010b), Spatial variabilities and their relationships of the trends of temperature, water vapor, and precipitation in the North American Regional Reanalysis, J. Geophys. Res., 115, D06110, doi:10.1029/2009JD013192. Lu, E., E. S. Takle, and J. Manoj (2010), The relationships between climatic and hydrological changes in the upper Mississippi River basin: A SWAT and multi-GCM study, J. Hydrometeorol., 11, 437–451.


12,586


10.1002/2014JD022456

Lu, E., Y. Luo, R. Zhang, Q. Wu, and L. Liu (2011), Regional atmospheric anomalies responsible for the 2009–2010 severe drought in China, J. Geophys. Res., 116, D21114, doi:10.1029/2011JD015706. Lu, E., et al. (2014), The atmospheric anomalies associated with the drought over the Yangtze River basin during spring 2011, J. Geophys. Res. Atmos., 119, 5881–5894, doi:10.1002/2014JD021558. Lv, Z., Y. Zou, and Y. Li (2011), Analysis on the rainstorm probability in the eastern region of Sichuan based on Poisson distribution, J. Chengdu Univ. Info. Tech., 25, 531–535. Michael, T. B., and L. O. Frederick (2011), Exploratory regression analysis: A tool for selecting models and determining predictor importance, Psychon. Soc., 8, 93–98. Mo, K. C., and E. H. Berbery (2011), Drought and persistent wet spells over South America based on observations and the U.S. CLIVAR Drought Experiments, J. Clim., 24, 1801–1820. Nandintsetseg, B., J. S. Greene, and C. E. Goulden (2007), Trends in extreme daily precipitation and temperature near lake Hövsgöl, Mongolia, Int. J. Climatol., 27, 341–347. Peña, M., and M. W. Douglas (2002), Characteristics of wet and dry spells over the Pacific side of central America during the rainy season, Mon. Weather Rev., 130, 3054–3073. Piccarreta, M., A. Pasini, D. Capolongo, and M. Lazzari (2013), Changes in daily precipitation extremes in the Mediterranean from 1951 to 2010: The Basilicata region, southern Italy, Int. J. Climatol., 33, 3229–3248. Pratt, J. W. (1987), Dividing the indivisible: Using simple symmetry to partition variance explained, in Proceedings of Second Tampere Conference in Statistics, vol. 11, edited by T. Pukilla and D. Duntaneu, pp. 245–260, University of Tampere, Tampere, Finland. Ratan, R., and V. Venugopal (2013), Wet and dry spell characteristics of global tropical rainfall, Water Resour. Res., 49, 3830–3841, doi:10.1002/ wrcr.20275. Reiser, H., and H. Kutiel (2012), The dependence of the annual total on the number of rain-spells and their yield in the Mediterranean, Geografiska Ann. Ser. A, 94, 285–299. Rodrigo, F. S. (2010), Changes in the probability of extreme daily precipitation observed from 1951 to 2002 in the Iberian Peninsula, Int. J. Climatol., 30, 1512–1525. Schmidli, J., and C. Frei (2005), Trends of heavy precipitation and wet and dry spells in Switzerland during the 20th century, Int. J. Climatol., 25, 753–771. Singh, N., and A. Ranade (2010), The wet and dry spells across India during 1951–2007, J. Hydrometeorol., 11, 26–45. Stocker, T. F., and C. C. Raible (2005), Water cycle shifts gear, Nature, 434, 830–833. Sun, Y., S. Solomon, A. Dai, and R. W. Portmann (2007), How often will it rain?, J. Clim., 20, 4801–4818. Suppiah, R., and K. J. Hennessy (1998), Trends in total rainfall, heavy rain events and number of dry days in Australia, 1910–1990, Int. J. Climatol., 18, 1141–1164. Tank, A. M. G. K., and G. P. Können (2003), Trends in indices of daily temperature and precipitation extremes in Europe, 1946–99, J. Clim., 16, 3665–3680. Teixeira, M. S., and P. Satyamurty (2011), Trends in the frequency of intense precipitation events in southern and southeastern Brazil during 1960–2004, J. Clim., 24, 1913–1921. Timm, O. E., H. F. Diaz, T. W. Giambelluca, and M. Takahashi (2011), Projection of changes in the frequency of heavy rain events over Hawaii based on leading Pacific climate modes, J. Geophys. Res., 116, D04109, doi:10.1029/2010JD014923. Villarini, G., J. A. Smith, and G. A. Vecchi (2013), Changing frequency of heavy rainfall over the central United States, J. Clim., 26, 351–357. Wang, D., L. Chen, and Y. Ding (2006), The trend in rainfall and wet days of China in recent 40 years and the correlation between the trend and the change of globe temperature, J. Trop. Meteorol., 22, 283–289. Wang, S., et al. (2000), Seasonal precipitation series of eastern China since 1880 and the variability, Acta Geogr. Sin., 55, 281–293. Wang, X., and P. Zhai (2008), Changes in China’s precipitation in various categories during 1957–2004, J. Trop. Meteorol., 24, 459–466. Wang, Y., and L. Zhou (2005), Observed trends in extreme precipitation events in China during 1961–2001 and the associated changes in large-scale circulation, Geophys. Res. Lett., 32, L09707, doi:10.1029/2005GL022574. Wu, P., R. Wood, and P. Stott (2005), Human influence on increasing Arctic river discharges, Geophys. Res. Lett., 32, L02703, doi:10.1029/ 2004GL021570. Yang, F., A. Kumar, M. E. Schlesinger, and W. Wang (2003), Intensity of hydrological cycles in warmer climates, J. Clim., 16, 2419–2423. Yu, X., and Y. Ding (2006), Probabilistic characteristics of rainstorm in eastern area of China: Poisson distribution-based statistical simulation, J. Nat. Disaster, 15, 13–18. Zhai, P., X. Zhang, H. Wan, and X. Pan (2005), Trends in total precipitation and frequency of daily precipitation extremes over China, J. Clim., 18, 1096–1108. Zhou, T., R. Yu, H. Chen, A. Dai, and Y. Pan (2008), Summer precipitation frequency, intensity, and diurnal cycle over China: A comparison of satellite data with rain gauge observations, J. Clim., 21, 3997–4010. Ziegler, A. D., et al. (2003), Detection of intensification in global- And continental-scale hydrological cycles: Temporal scale of evaluation, J. Clim., 16, 535–547. Zolina, O., C. Simmer, K. Belyaev, A. Kapala, and S. Gulev (2009), Improving estimates of heavy and extreme precipitation using daily records from European rain gauges, J. Hydrometeorol., 10, 701–716. Zolina, O., C. Simmer, K. Belyaev, S. K. Gulev, and P. Koltermann (2013), Changes in the duration of European wet and dry spells during the last 60 years, J. Clim., 26, 2022–2047.

LU ET AL.


12,587