Statistical tests for a correlation between decadal variation in June precipitation in China and sunspot number

Authors

  • Jing-Song Wang,

    1. National Center for Space Weather, National Satellite Meteorological Center, China Meteorological Administration, Beijing, China
    Search for more papers by this author
  • Liang Zhao

    Corresponding author
    1. National Center for Space Weather, National Satellite Meteorological Center, China Meteorological Administration, Beijing, China
    2. National Climate Center, Beijing, China
      Corresponding author: L. Zhao, National Center for Space Weather, National Satellite Meteorological Center, China Meteorological Administration, 46 Zhongguancun South St., 100081 Beijing, China. (zhaol@cma.gov.cn)
    Search for more papers by this author

Corresponding author: L. Zhao, National Center for Space Weather, National Satellite Meteorological Center, China Meteorological Administration, 46 Zhongguancun South St., 100081 Beijing, China. (zhaol@cma.gov.cn)

Abstract

[1] Six different statistical methods (i.e., correlation, difference, prominent period, variance contribution, scale-averaged spectrum, and cross spectrum) are used to test for regional differences in the relationship between the 11 year sunspot cycle and June precipitation in China during the 20th century. In the Huaihe River basin (HRB) of central China, located at the marginal region of the East Asian summer monsoon (EASM), there exists a reliable positive-correlation relationship between the 11 year sunspot cycle and June precipitation; whereas, possible negative and very weak positive correlations in the south of the middle–lower Yangtze River Region and the HeTao Basin (HTB), located in the interior of the EASM and the westerlies, respectively. The reasons for these regional differences are investigated, revealing that the marginal region of EASM may be more sensitive to solar forcing than is its interior, which results in the HRB becoming the most susceptible (strongest correlation) region. That is to say, in June during the high sunspot number (SSN) years, the influence of the EASM is significantly greater and more to the north than that in June during the low SSN years, causing the HRB to be mainly influenced by the EASM (westerlies) in June during the high (low) SSN years. The northward expansion of the June EASM probably resulted from enhancement of the low-level southwesterly monsoon flow over the northern tropical Indian Ocean, combined with an expansion of the western Pacific subtropical high at times of high SSN.

1. Introduction

[2] Many decadal climate phenomena related to precipitation are often related empirically to the 11 or 22 year solar cycle (SC) activity [e.g., Hoyt and Schatten, 1997; Herman and Goldberg, 2005; Zhao et al., 2011], even though the cause-and-effect connection remains obscure. However, the complex feedbacks and variability of a regional climate system can make solar signals ambiguous. Consequently, on the regional scale, the relations between precipitation and the SC are rather complex and diverse. The drought cycle in North America [Stockton et al., 1983; Perr, 1994; Currie, 1996b; Cook et al., 1997]; precipitation in Africa [Currie, 1993, 1996b], Australia, and South America [Currie, 1996a; Thresher, 2002]; and monsoon precipitation in India and Arabia [Kodera, 2004; Bhattacharyya and Narasimha, 2005; Lihua et al., 2007] all show some correlation with variations in solar activity, especially on decadal or two-decadal scales. However, some of the correlations are positive and some are negative; opposite correlations are even found in adjacent regions [e.g.,Thresher, 2002; Meehl et al., 2009; Zhao et al., 2012]. Although higher spatial and temporal resolutions are required to improve the accuracy and consistency of some correlations, or the time period of available data needs to be extended to establish statistical significance, the fact that opposite correlations are found in adjacent regions may suggest that in the interior of a regional climate system (e.g., the monsoon systems, subtropical highs, etc.), the solar signal as an external forcing can be modified by its feedbacks and other physical processes. However, near the margins of a regional climate system, the solar signal may be stronger than in the interior, e.g., the observed quasi-decadal signals at the northern margins of the Antarctic circumpolar vortex [Thresher, 2002] and the East Asian summer monsoon (EASM) [Zhao et al., 2012], or the solar signal may be also more unreliable in the regions with lots of variability than in the interior.

[3] Previous studies have reported that natural decadal variability likely dominates summer precipitation in the monsoon region of China [e.g., Lei et al., 2011], and some SC signals in central and east Asia have been detected in the variability of precipitation deduced from historical documents or paleoclimate proxy records [e.g., Currie, 1995, 1996b; Davi et al., 2006; Tan et al., 2008]. Although Zhao et al. [2012] reported three regions (the HeTao Basin (HTB), the Huaihe River Basin (HRB), and to the south of the middle–lower Yangtze River Region (SYR)) in China with a strong correlation between sunspot number (SSN) and June precipitation, with the correlation being positive/negative in regions north/south of the Yangtze River (see Figure 1b or Zhao et al. [2012, Figure 1b]), they have not tested the validity of the strong correlations obtained for these three regions. Furthermore, a possible physical reason for the correlations has not been studied in detail. On the other hand, the regional climate systems that dominate the climate in these strong-correlation regions in summer are usually different in each region. Therefore, it need to be confirmed whether there is a significant regional response difference in these regions and be established what is the physical relation between the possible regional difference and regional climate systems (e.g., the EASM, the western Pacific subtropical high (WPSH)). Since the intensity anomaly of the EASM is usually a direct cause of summer drought and flooding in China [Zhu, 1934], it should be a key factor in examining whether and how the influence of the 11 year SC on regional climate is amplified by the regional monsoon system or by the other systems in the past century. Therefore, it would be beneficial to identify the possible influence of the monsoon on the relation between the sun and precipitation, to test for regional differences in precipitation between the monsoon margin, interior, and nonmonsoon regions of China.

[4] In the present study, a current 106 year-long high-resolution record of land-surface precipitation, based on observations, is analyzed by statistical and mathematical methods to test for a connection between the 11 year SC and decadal precipitation variations in various regions of China. We focus on the onset of the EASM, which is assessed for correlations with both statistical and physical significance, and we examine the causes of regional differences in precipitation in response to solar variability. We also examine the relationship between longitudinal variations in the monsoon and the SSN.

2. Data and Selection of Temporal-Spatial Coverage

[5] The monthly land-surface precipitation data used in this study are CRU TS 3.0 with 0.5° × 0.5° resolution across 89.75°S–89.75°N and 0.25°E–359.75°E from 1901 to 2006, compiled by the University of East Anglia Climatic Research Unit (CRU) [Mitchell and Jones, 2005]. CRU TS 3.0 are based on observations and are long enough to study decadal and interdecadal precipitation variations during the past 106 years; i.e., 10 SCs. However, in the first half of the 20th century, there are few stations for precipitation in China in CRU TS 3.0 data. Thus, another data set, monthly accumulated precipitation of 160 land stations in China from 1951 to 2011, is used to confirm the results based on CRU data. This data set comes from Beijing Climate Center (BCC). Wind data are from the U.S. National Oceanic and Atmospheric Administration–Cooperative Institute for Research in Environmental Sciences (NOAA-CIRES) 20th Century Reanalysis version 2 Monthly Averages (1871–2008) with a spatial resolution of 2.0° for 1000 to 10 hPa [Compo et al., 2011]. The yearly mean relative SSN, used as the primary proxy of solar activity, is from the U.S. NOAA National Geophysical Data Center (NGDC) (ftp://ftp.ngdc.noaa.gov/STP/SOLAR_DATA). The annual AO index is obtained from Li and Wang [2003].

[6] The 106 year (1901–2006) mean monthly CRU precipitation rates (monthly mean relative to annual mean precipitation), averaged between 26.25°N and 34.25°N (where the EASM margin at 700 hPa in June usually lies) along 115.25°E (across three different climate regions outlined in Figure 1b) during 1901–2006, is shown in the top part of Figure 1a. The mean precipitation rate is highest in June (∼15%) among 12 months, during the peak rainy season in the middle–lower Yangtze River region and nearby areas, which are typical monsoon regions, where summer precipitation is usually much higher than winter precipitation.

Figure 1.

(a) (top) The 106 year (1901–2006) mean monthly CRU precipitation rates (monthly mean relative to annual mean precipitation) (bar diagram, units: %) averaged between 26.25°N and 34.25°N along 115.25°E; (bottom) the time-latitude section along 115.25°E of significant linear correlations (shaded areas) between 8 year low-pass filtered monthly CRU precipitation and unfiltered corresponding annual SSN for the period 1901–2006. (b) Significant linear correlations (shaded areas) between 8 year low-pass filtered June CRU precipitation at each grid and unfiltered corresponding annual SSN for the period 1901–2006, and unfiltered June 700 hPa horizontal wind (vector arrows, units: 10 m/s) averaged over the 106 years. (c) The same as Figure 1b, but for the period 1951–2011 using 8 year low-pass filtered BCC precipitation data from 160 stations in China. (d) Vertical meridional circulation (vector arrows) along 110°E cross-section of unfiltered June horizontal meridional wind (units: 10 m/s) and vertical pressure velocity (units: −0.02 Pa/s, so that positive is upward), averaged over the 106 years. In Figures 1a–1c, a Monte Carlo method is used to establish confidence intervals for correlations of the filtered data, and dark and light gray shading denotes significant (at the 95% confidence level; i.e., absolute values of correlations above 0.38 and 0.50 for Figures 1a and 1b and Figure 1c, respectively) positive and negative correlations, respectively. The top, middle, and bottom rectangles in Figures 1b and 1c and the right, middle, and left rectangles in Figure 1d outline HTB, HRB, and SYR, respectively. Thick contours in Figures 1b and 1d are mean southerly velocity contours of 0 m/s, representing the margin of the EASM. For clarity, only half of the arrows along the longitude and latitude direction are shown in Figure 1b.

[7] To remove the dominance of the higher-frequency components in the precipitation data (e.g., the components probably influenced by El Niño–Southern Oscillation (ENSO) with a typical period of 2–7 years), signals with periods of less than 8 years in the precipitation data are filtered out by a low-pass filter using fast Fourier transform (FFT). The bottom part ofFigure 1ashows a time-latitude section along 115.25°E of significant (95% confidence) correlations between the 8 year low-pass filtered monthly precipitation of the CRU data and unfiltered annual SSN during the 106 years. The confidence intervals are different for the unfiltered and filtered data, so a Monte Carlo test is used to define the confidence intervals [cf.Zhao et al., 2012]. The threshold value of correlation coefficients at the 95% confidence level turns out to be 0.38. March and June are the months with the most significant correlations during the spring (March, April, May) and summer period (June, July, August), respectively, and in June we observe regions with significant positive and negative correlations simultaneously. This paper focuses on these regions during the summer strong-correlation period (i.e., June), and attempts to explain the origin of the positive and negative correlations.

[8] The spatial distributions of the correlation (shaded) between the SSN and 8 year low-pass filtered June precipitation at each grid inFigure 1b (using CRU precipitation data for the period 1901–2006) and c (using BCC precipitation data for the period 1951–2011) both show significant (95% confidence) correlations (above 0.38 for the 106 years and 0.50 for the 61 years, respectively) in the following representative regions of China (indicated by rectangles in Figures 1b and 1c): the HTB (106°–116°E, 38°–45°N), the HRB (105°–120°E, 31°–35°N), and the SYR (110°–120°E, 25°–29°N). In Figure 1c, the Cressman interpolation technique [Cressman, 1959] is used to interpolate the correlation coefficients between the 160 irregular station data and SSN to a 0.5° × 0.5° grid. The significance of the correlation fields needs to be evaluated. Monte Carlo simulations are used to determine whether the observed significant correlation patterns are greater in size than those expected by chance [cf. Livezey and Chen, 1983]. By correlating a random SSN time series (from its empirical distribution) with the time series of 8 year low-pass filtered June precipitation at the 14113 land-surface grid points across 0.25°N–60.25°N and 50.25°E–150.25°E forFigure 1b or at the 160 stations in China for Figure 1c in 500 different simulations, an accurate probability density function of the number of significant grid points or stations in each simulation was estimated. As a result, the percentage that the absolute values of correlations exceed 0.38 for Figure 1b or 0.50 for Figure 1c at the 95% confidence level turns out to be 12.22% (1727 grid points) for Figure 1b or 18.75% (30 stations) for Figure 1c, and the correlation field is considered significant at the 80% (for Figure 1b) or 99.9% (for Figure 1c) confidence level. The results show that the confidence levels of the two correlation fields of Figures 1b and 1c are very different. The reason is likely there are few stations for precipitation in the CRU data especially in western China during the first half of the 20th century, which reduces the confidence of the correlation field. Thus the confidence test of the correlation field of Figure 1c using the BCC 160 station data should be more reliable than that of Figure 1b using the CRU grid data.

[9] The arrows in Figures 1b and 1dshow the circulations averaged over the 106 years. The northward and southward winds originate in the monsoon region and the westerlies, respectively. Therefore, the north margin of the EASM might well be denoted by the main 0 m/s contour of southerly velocity near the 30°–40°N. So it can be seen that in June, these three high-correlation regions are just influenced by different climatological mean circulations, respectively. The HTB is dominated by drier and colder flow from the northwest at the middle and upper troposphere, the SYR by wetter and warmer monsoon flow from the southwest at the mid and lower troposphere, and the HRB is located near the boundary between these circulations. So selecting June is of benefit to testing the difference between the margin and interior of the EASM and the westerlies. Therefore, these three regions are chosen as the main regions for analysis, and June as the main period, as this is when changes in precipitation, both in space and time, are significantly correlated with the SSN.

3. Tests of the Relationship Between the Sunspot Cycle and Precipitation in Different Regions of China

3.1. Correlation Test

[10] Correlation between annual SSN and the unfiltered, 8 year low-pass filtered monthly (annual) precipitation is shown inFigures 2a–2c. The highest positive correlation (for the HTB and HRB) and the lowest negative correlation (for the SYB) all occur in June (confidence level >95%), and low-pass filtering results in an enhanced correlation in nearly every month of the year. In June, using unfiltered data, the correlation coefficients in all three regions are <0.3. The top parts ofFigures 2d–2fshow that precipitation is generally coherent with the SSN. The higher-frequency components in precipitation may represent noise, which reduces the correlation coefficient. Thus, the scenarios after filtering are much more evident than before filtering (see the bottom parts ofFigures 2d–2f). In the HTB, precipitation shows a change from being primarily antiphase with the SSN during the period before the 1940s (long dashed frame in Figure 2d) to being in phase after the 1940s. Therefore, we consider that the correlation in the HTB is uncertain. What caused the sign change of correlation between June precipitation in the HTB and SSN? Chen and Zhou [2012] indicated a solar impact on the winter climate in North China via the Arctic Oscillation (AO). So we investigated the correlation between the annual AO index from Li and Wang [2003] and the SSN, and found that the AO index decreased abruptly in early 1940s, and the correlations between the annual AO index and the SSN changed from −0.32 (96% confidence) for the period 1901–1940 to 0.11 (63% confidence) for 1941–2006. Therefore, the sign change of correlation between June precipitation in the HTB and SSN may be related to the AO. In the other two regions, the correlation is consistently of the same sign during the entire 106 years.

Figure 2.

(a–c) Correlation coefficients between the unfiltered/8 year low-pass filtered, monthly precipitation, and corresponding annual SSN for the period 1901–2006 (horizontal short/long dashed lines denote 95% confidence level lines for unfiltered/filtered data using a Monte Carlo test; vertical dashed line indicates June; the abscissa ‘Y’ denotes the annual scenario). (d–f) Standardized time series of unfiltered (top parts of plot)/8 year low-pass filtered (bottom parts), regional June precipitation, and corresponding annual SSN. HTB for Figures 2a and 2d, HRB for Figures 2b and 2e, SYR for Figures 2c and 2f.

3.2. Test of Significant Difference

[11] High-solar-activity years (HSY) or low-solar-activity years (LSY), corresponding to the maximum or minimum SSN and 1 year either side in each SC for the period from 1901 to 2006, are chosen (Table 1). Figure 3shows the difference between HSY and LSY for the 8 year low-pass filtered monthly precipitation in the three regions. Because the 8 year low-pass filtering of precipitation data has an effect on its statistical significance of difference, a Monte Carlo test is carried out to define the confidence intervals. According to Monte Carlo method, a random time series (106 samples in each series) is generated and is 8 year low-pass filtered using FFT. And two groups (30 and 29 samples are chosen from the series according toTable 1, respectively) are established. Then a t statistic is calculated to test whether there is a significant mean difference between two groups. These steps are repeated for 100,000 times. Finally, according to sort of absolute values of these t statistics, threshold values of t statistic and difference at 95% confidence level are defined. Figure 3 shows that in both the HRB and the SYR, the month with the maximum difference is June, with a significant difference (at about 95% confidence level); in contrast, in the HTB the greatest difference is seen in July (not statistically significant). The mean HSY (LSY) June precipitation in the HTB, the HRB, and the SYR is 38 (33), 123 (102), and 235 mm (270 mm), respectively, with the HSY in the HTB and the HRB exceeding LSY by 16% and 21%, respectively, and the HSY in the SYR being 13% less than LSY. Thus, the difference testing reveals that in the HRB and the SYR, the HSY June precipitation is significantly different from the LSY precipitation, but this is not the case in the HTB.

Table 1. HSY or LSY Over the Period 1901–2006a
HSY (30)LSY (29)
  • a

    The numbers in the parentheses are sample numbers. Because CRU TS 3.0 data starts from 1901 that is just the minimum year of SSN, the LSY in this SC (1901–1912) can be chosen are only 1901 and 1902.

1904, 1905, 1906, 1916, 1917, 1918, 1927, 1928, 1929, 1936, 1937, 1938, 1946, 1947, 1948, 1956, 1957, 1958, 1967, 1968, 1969, 1978, 1979, 1980, 1988, 1989, 1990, 1999, 2000, 20011901, 1902, 1912, 1913, 1914, 1922, 1923 1924,1932, 1933, 1934, 1943, 1944, 1945, 1953, 1954, 1955, 1963, 1964, 1965, 1975, 1976, 1977, 1985, 1986, 1987, 1995, 1996, 1997
Figure 3.

Difference (bar; units: mm) in composite mean monthly precipitation between HSY and LSY after 8 year low-pass filtering (statistical significance was calculated using thet test). Long dashed lines denote the 95% confidence level using a Monte Carlo test, and the vertical dashed line denotes June. (a) HTB, (b) HRB, and (c) SYR.

3.3. Prominent Period Test

[12] When using wavelets for feature extraction, Morlet's wavelet is a suitable choice because it provides a good balance between time and frequency localization [Grinsted et al., 2004]. Here, wavelet power spectral analysis [cf. Torrence and Compo, 1998] with Morlet's wavelet as the mother wavelet is applied to regional precipitation as well as the SSN. Figure 4 shows the global wavelet spectra of the unfiltered regional June precipitation and the corresponding annual SSN. None of the precipitation spectral peaks with a period greater than 8 years passes the 95% confidence test. The precipitation spectral peak at 11.3 years for the HRB is at about the 80% confidence level, which is the highest confidence for all spectral peaks with a period greater than 8 years in all three regions.

Figure 4.

Global wavelet spectra of unfiltered annual SSN (thick solid lines) and the unfiltered regional June precipitation (thick dashed lines, units: mm2) using Morlet's wavelet as the mother wavelet. The thin solid (dashed) lines denote 95% confidence levels for the global wavelet spectra of SSN (precipitation). (a) HTB, (b) HRB, and (c) SYR.

3.4. Variance Contribution Test

[13] Variance contribution can show the degree to which decadal variations in precipitation contribute to total long-period variability. We calculated the variance contribution of the 9–13 year component to the total and long-period components of the precipitation data. In the HRB, the variance contribution is the highest, and the 9–13 year signals contribute >15% to the total variance of unfiltered precipitation (figure not shown) and >50% to precipitation low-frequency (>8 year time scale) variance in most regions of the HRB (Figure 5). In the other two regions, the variance contributions are much lower. This result demonstrates that for precipitation in the HRB of central China, the decadal component is the largest among low-frequency signals, and it dominates the long-period (>8 years) variation.

Figure 5.

Variance contribution (units: %) of the 9–13 year component of June precipitation in China for the long-period (>8 years) portion during 1901–2006. The top, middle, and bottom rectangles outline HTB, HRB, and SYR, respectively.

3.5. Scale-Averaged Spectrum Test

[14] The scale-averaged wavelet power (SAWP) is a time series of the average variance in a certain band, which can be used to examine the modulation of one time series by another [Torrence and Compo, 1998]. SAWP may help us detect the modulation by assessing the temporal behavior of the spectral power of precipitation and the SSN, and their relationship to each other. Figure 6shows the temporal evolution of the SAWP over the 9–13 year band for the annual SSN and the June unfiltered, 8 year low-pass filtered precipitation in the three regions. In the HTB and HRB, the wavelet power spectra of precipitation tend to track that of the SSN (correlation coefficient of 0.74 and 0.92). And the filtered precipitation shows significantly (95% confidence level) higher power during the1950s–1960s for the HTB and during the 1940s–1990s for the HRB, suggesting that during the epochs the precipitation in the HTB and HRB has a strong quasi 11 year oscillation. The quasi 11 year power of the SSN is also strong during this epoch. In contrast with the HTB and the HRB, the 9–13 year SAWP in the SYR seems to be in antiphase with that of the SSN (correlation coefficient of −0.55), and the period with the strongest 11 year period is distinct from that of the SSN. Thus, we can speculate that the response region, where precipitation is modulated by the 11 year sunspot cycle, may have drifted in the 1940s from the south to the north of the Yangtze River, which is coincident with the increase in SSN.

Figure 6.

Temporal evolutions of the scale-averaged Morlet's wavelet power over the 9–13 year band for (top) unfiltered and (bottom) 8 year low-pass filtered regional June precipitation and corresponding annual SSN. Horizontal solid and dashed test lines denote 95% confidence levels for SSN and precipitation, respectively. (a) HTB, (b) HRB, and (c) SYR.

3.6. Cross-Spectrum Test

[15] The continuous wavelet transform (CWT) is a common tool for analyzing localized intermittent oscillations in a time series. The cross-wavelet transform (XWT) constructed from two CWTs can reveal their common power and relative phase shift [Torrence and Compo, 1998; Grinsted et al., 2004]. If the precipitation in the three regions of the present study is related to the sunspot cycle, their common power and a consistent or small phase lag should be detectable. The XWTs of the SSN and June precipitation in the three regions are shown in Figure 7, showing a significant continuous common power in the 8–14 year band, especially in the HRB and SYR. The relative phase relationship (arrows in Figure 7) shows that in the sectors with significant common power (near the 11 year band), in the HTB the SSN is in phase with or slightly lags precipitation post-1940, but is antiphase with precipitation pre-1940. In the HRB, the SSN is primarily in phase with or slightly leads the precipitation. In the SYR, the SSN is primarily antiphase with the precipitation. This result indicates that throughout most of the 20th century, in the HRB and SYR, the decadal oscillations are nearly phase locked, whereas in the HTB the relationship between the decadal oscillation of precipitation and the SSN is suspect.

Figure 7.

Cross-wavelet transforms of unfiltered regional June precipitation and corresponding annual SSN. The 5% significance level against red noise is shown as a thick contour. The relative phase relationship is shown by arrows (with in-phase pointing right, antiphase pointing left, and SSN leading precipitation by 90° pointing straight down). (a) HTB, (b) HRB, and (c) SYR.

3.7. Overall Evaluation and Confirmation

[16] The above analysis shows 18 results from 6 different methods (i.e., correlation, difference, prominent period, variance contribution, scale-averaged spectrum, and cross spectrum) for 3 regions in China, each with its own advantages and disadvantages. These test results are possibly inconsistent. In order to better evaluate the reliability of the relationship between SSN and precipitation for various regions in China, evaluation scores were made according to the present test results (Table 2). The symbols +/− denote passing/not passing the corresponding test (scoring +1/−1 point), and 0 denotes uncertainty; the values in the bottom row are total scores. The HRB scores the highest with a perfect score (4 points), indicating that the relationship between the SSN and precipitation in the HRB is the most reliable of the three regions. The SYR scores 0 points, suggesting that the relationship is weak. The HTB scores −2 points, implying that there is no apparent relationship. Therefore, we can be relatively certain that in central China (i.e., in the HRB) the June precipitation is positively correlated with the 11 year SC. In contrast, a negative correlation may exist in southern China.

Table 2. Evaluation Scores for the Reliability of the Relationship Between the SSN and Precipitation in the HTB, HRB, and SYRa
TestsRegions
HTBHRBSYR
  • a

    Here, +/− represents passing/not passing the corresponding test (scoring +1/−1 points), and 0 represents uncertainty (scoring 0 points), according to the results of the previous analysis.

Correlation0++
Difference-++
Period---
Variance contribution-+-
Scale-averaged spectrum++-
Cross spectrum0++
Total scores−240

4. Discussion

[17] The three regions examined in this study show differences in terms of the dominant regional climate factors. Figure 1b shows that in June the summer monsoon usually affects southern China, the Yangtze River basin, and part of the HRB (the domains influenced by the EASM at 700 hPa might well be denoted by the southerly wind regions). During this period, the HRB usually lies on the northern boundary of the EASM, whereas the HTB has yet to be influenced by the EASM. However, as shown in Figure 8a, during the HSY and LSY (see Table 1), the influenced domains by the EASM are significantly different, especially in the HRB, where the EASM is much larger during the HSY than during the LSY at 700 hPa, with a significant southerly difference flow. That is, during the HSY, the HRB is mainly influenced by (and the HTB is partially influenced by) the EASM, whereas during the LSY they are slightly influenced or not influenced by the EASM. Moreover, Figure 8a shows that the prevailing significant southwesterly difference winds are nearly continuous from Bengal to southwest China, HRB, and HTB (the distribution of disconnected areas is probably influenced by terrain), suggesting a stronger summer monsoon flow, which is inclined to cause rain during the HSY. However, the anomalous anticyclonic difference circulations over the central regions of the EASM (specifically over the SYR and the south of Japan), with a hint of weaker monsoon flow, do not favor precipitation during the HSY. Further this relation is given by the circulation difference in the vertical cross section along 110°E (Figure 8b). It is clear that the 0 m/s contours of southerly velocity are generally located more to the north and east during the HSY than during the LSY, with significant northward and upward difference flow to the north of 30°N over the HRB. This finding indicates that the influenced domain by the EASM during the HSY is significantly different from that during the LSY, when it rains more heavily than during the LSY in the HRB. However, over the SYR, downwelling anomalies limit convection and precipitation, producing drier conditions. This result supports the previous conclusion that the margin of a regional climate system is probably more sensitive to external forcing (e.g., solar forcing) than its interior.

Figure 8.

Composite difference map (vector arrows) of June wind between HSY and LSY for the period 1901–2006, showing (a) horizontal wind (units: 1.2 m/s) at 700 hPa constant pressure surface and (b) meridional wind (units: 1.2 m/s) and vertical pressure wind (units: −0.02 Pa/s, so that positive is upward) along longitude 110°E cross section. The thick solid/long dashed contours indicate the HSY/LSY mean southerly velocity contours of 0 m/s, representing the margin of the EASM. Darker and lighter shaded areas indicate regions where the difference in meridional wind is significant at the 95% and 80% confidence levels (t test), respectively. The top, middle, and bottom dashed line rectangles in Figure 8a and the right, middle, and left dashed line rectangles in Figure 8b outline HTB, HRB, and SYR, respectively. For clarity, only half of the arrows along the longitude direction are shown in Figure 8a. The solid line box (105°E–120°E, 25–45°N) in Figure 8a outlines the region where the grid number of southerlies (GNS) is used to define a relative size of domain area of EASM in China.

[18] With regard to the direct cause of the variation in the location of the northern boundary of the EASM, the monsoon flow over China originates mainly in the tropical Indian Ocean and the western Pacific Ocean, and blows northward to China when affected by the WPSH. In general, the EASM is influenced by both the monsoon flow over the tropical Indian Ocean and the WPSH. When the monsoon flow over the tropical Indian Ocean is stronger and the WPSH is larger (more to the north or west), the northern boundary of the EASM will probably shift poleward. Qun and Qiuming [1993] used monthly mean area index (MAI) to measure the area of subtropical high and pointed out a lag in the subtropical high responses to solar forcing during 1954–1991. The MAI for the WPSH (WPSHI) is the sum of grid points exceeding 5880 gpm at 500 hPa across 110°E–180°E and 10°N–90°N. Herein, the WPSHI is used to give quantitative assessments for the area of WPSH. And the north boundary position of the EASM also needs a quantitative assessment. Because the grid number of southerlies (GNS) at 700 hPa over the studied region (105°E–120°E, 25°N–45°N) (solid line box in Figure 8a) in June can generally indicate a relative size of domain area of June EASM in China and indirectly reflect a relative meridional position of the north boundary of June EASM in China, it is used here to measure the EASM. Figure 9 shows the time series of the June WPSHI, the June GNS and the annual SSN from 1901 to 2006, and the correlations between them are shown in Table 3. The confidence intervals are different for the unfiltered and filtered data, so a Monte Carlo test is carried out to define the confidence intervals [cf. Zhao et al., 2012]. Table 3 shows that the positive correlations all exist between them during the 106 years. This suggests that in June of the high SSN years, the domain area of the EASM is likely larger and more to the north, and the WPSH is also larger, than those in the low SSN years. Moreover, in June of the HSY, a significantly stronger monsoon flow over the Bay of Bengal region can be clearly observed (Figure 8a). The stronger monsoon flow over the Bay of Bengal region and the larger WPSH, are likely to explain the northward expansion of the June EASM during the HSY.

Figure 9.

The time series of the unfiltered (dashed lines)/8 year low-pass filtered (thick solid lines) June 500 hPa WPSHI (top curve) across 110°E–180°E and 10°N–90°N, June 700 hPa GNS (bottom curve) across 105°E–120°E and 25°N–45°N (solid line box inFigure 8a) and unfiltered annual SSN (shaded) from 1901 to 2006.

Table 3. The Correlations Between Unfiltered/8 Year Low-Pass Filtered June WPSHI, GNS, and Unfiltered Annual SSNa
 WPSHIGNS
  • a

    Percentage in bracket denotes confidence level obtained based on Monte Carlo tests.

SSN0.26 (99%)/0.36 (94%)0.20 (96%)/0.31 (89%)
WPSHI 0.17 (92%)/0.50 (99%)

[19] Overall, the domain of influence of the EASM is likely modulated by the 11 year SC, leading directly to the migration of the summer monsoon rainband in China. The domain of the EASM is presumably directly related to the intensity of low-level southwesterly monsoon flow over the Bay of Bengal region and to the area of the WPSH. That is, solar activity is likely to have an indirect influence on June precipitation in China via the magnifying or diminishing role of the EASM, and the decadal signals of EASM can be traced back to the Indian monsoon region and the western Pacific. However, it remains unclear how these regional climate systems are modulated by solar activity on the decadal time scale.

[20] Previous studies have reported that the mode associated with the solar sunspot cycle is dominant and may amplify the variability at higher levels of the atmosphere, providing support for a mechanism through which solar influence is transmitted downward from the stratosphere, affecting climate [e.g., Haigh, 1996, 1999; Shindell et al., 1999; Dima et al., 2005; Ineson et al., 2011]. It can be seen from Figures 1d and 8bthat significant differences in meridional wind between high- and low-solar June exist at the top and bottom of the EASM circulation, as well as to the south and north. These significant differences exist mainly near the margins or boundaries of some regional climate systems. However, further studies are required to determine whether the forcing of the decadal, meridional variation in circulation of the monsoon and the WPSH is derived from above or below (or both) or elsewhere, and whether the main forcing depends on the season. In addition, more research is required to determine whether the decadal meridional shift of the rain belt is actually a reflection of the onset time (early or late) of the monsoon. Nevertheless, it should be noted that in the complicated chain of events that occur from the sun to the earth's surface, the monsoon system probably plays a critical bridging role between the sun and surface precipitation, and may amplify solar signals or respond to amplified signals.

5. Conclusion

[21] This paper investigated regional differences in the relationship between solar variability and decadal June precipitation in China, and their causes, by analyzing monthly high-resolution land-surface precipitation data from 1901 to 2006 that are associated with sunspot number data and wind data using NOAA-CIRES 20th Century Reanalysis version 2 products. By selecting three regions with strong correlations and calculating the correlation coefficients, significant difference, prominent period, variance contribution, scale-averaged spectra and cross spectra, we tested the relationship between the 11 year sunspot cycle and June precipitation in China. The reliability of the relationship was also evaluated by using a method of scoring. In the HRB of central China, the relationship is most reliable; in the SYR, June precipitation may be negatively correlated with the SSN on >8 year time scales; in the HTB, in contrast, the relationship is weak.

[22] The direct cause of the regional difference in precipitation on decadal time scales is likely to be differences in the response of the EASM to the HSY and LSY, which likely makes the northern margins of the EASM (i.e., the HRB) the most sensitive region to variability in solar signals. Furthermore, the sun-EASM link is supported by a decadal variation characteristic of the monsoon. That is, in June, during the HSY, the influenced domains by the EASM are usually significantly larger than during the LSY, so that the monsoon flow influences most regions of the HRB and results in increased precipitation in such regions, and vice versa. Furthermore, the enhancement of the low-level southwesterly monsoon flow over the northern tropical Indian Ocean, and the expansion of the WPSH with high SSN, probably directly induce the northward expansion of the June EASM.

[23] These results confirm the occurrence of regional differences in precipitation among the monsoon margin, interior, and nonmonsoon regions of China during the special period (the EASM onset period), and confirm the occurrence of latitudinal and longitudinal variations in the solar signal in the monsoon and atmospheric circulation through the sunspot cycle. These factors can directly explain the regional differences, which suggests that amplifying or transferring solar signals should be related to the monsoon, but it is unknown whether the monsoon plays an amplifying or transferring role. Therefore, although these results are all statistically significant, further research is required to identify the factor(s) and process(es) that result in the significant differences in monsoon effect during the different SC phases and that induce anomalies in the circulation patterns, by distinguishing between signals of major climate regulators, climate response, and forcing.

Acknowledgments

[24] We thank the Climatic Research Unit (CRU) of the University of East Anglia for the precipitation data set and NGDC for the SSN data. Support for the Twentieth Century Reanalysis Project data set is provided by the U.S. Department of Energy, Office of Science Innovative and Novel Computational Impact on Theory and Experiment (DOE INCITE) program, and Office of Biological and Environmental Research (BER), as well as by the National Oceanic and Atmospheric Administration Climate Program Office. We are grateful to Jianyong Lu for his help. This research was supported by the National Natural Science Foundation (40931056) and the National Basic Research Program of China (2012CB957801 and 2012CB417205). Wavelet analysis software was provided by C. Torrence and G. Compo (http://paos.colorado.edu/research/wavelets/) and modified by A. Grinsted et al. (http://www.pol.ac.uk/home/research/waveletcoherence).

Advertisement