The variability of East Asian summer monsoon (EASM) precipitation is affected by anomalous lower boundary conditions, such as SST, sea ice, snow, and soil moisture ( Chang et al., 2000; Chen and Wu, 2000; Kripalani et al., 2002; Zhao et al., 2004; Ding and Chan, 2005; Zuo and Zhang, 2007; Wang et al., 2012). Among the above-mentioned boundary conditions, snow anomalies play an important role in modulating EASM precipitation. The memory of anomalous winter-spring snow in the climate system resides in the wetness of the underlying soil as snow melts during the spring and summer seasons. Therefore, the preceding snow anomaly has a lingering effect on the summer rainfall ( Shukla and Mooley, 1987; Bamzai and Shukla, 1999). There has been some research into the relationship between the preceding winter-spring snow anomaly over the Tibetan Plateau (TP) and the following spring or summer rainfall over East Asia ( Chen and Wu, 2000; Wu and Qian, 2000; Zhang et al., 2004; Wu and Kirtman, 2007; Zhao et al., 2007). This time-lag relationship may be explained by the role of soil moisture anomalies as a bridge linking the previous season’s snow over the TP and the subsequent EASM rainfall ( Zhao et al., 2007).

Although snow cover melts and disappears from winter and spring to summer over the main body of the eastern TP, it may persist through summer over the high altitudes, such as the western and southern parts of the TP where there are large mountain ridges ( Pu et al., 2007; Wu et al., 2012a). The simultaneous effects of summer snow cover over the TP have been found to be important for the decadal to interdecadal variations of heat wave over northern China and for the ENSO teleconnections ( Wu et al., 2012a, b). Recently, it has been found that the summer snow cover area proportion (SCAP) over the TP has a significant positive correlation with simultaneous precipitation over the Meiyu-Baiu region on the interannual time scale (Fig. 1, Liu et al., 2014). The potential mechanism for the high correlation is as follows. The SCAP anomaly has its independent effect and may directly modulate the land surface heating and consequently vertical motion over the western TP, and thus induce anomalous vertical motion over the North Indian Ocean via a meridional vertical circulation. Anomalous vertical motion over the North Indian Ocean may, in turn, result in an anomalous high over the western North Pacific and modulate the convective activity in the western Pacific warm pool, which stimulates the East Asia-Pacific (EAP) pattern ( Huang and Li, 1988; Huang and Sun, 1992) and eventu-ally affects summer precipitation over the Meiyu-Baiu region ( Liu et al., 2014).

Figure 1

	Figure Option ViewDownloadNew Window
	Figure 1 Spatial distribution of correlation coefficients between the summer snow cover area proportion (SCAP) index and summer precipitation for the period 1979-2006 with the shaded areas denoting correlation at the 90% confidence level ( Liu et al., 2014).

A previous study ( Liu et al., 2014) revealed the importance of the summer snow anomaly over the TP in modulating simultaneous precipitation over the Meiyu-Baiu region. However, several questions remain unclear and should be systemically discussed. For example, how persistent is the TP snow cover from the previous seasons to summer? Can summer rainfall over East Asia be forecasted using the preceding snow cover anomaly over the TP? In other words, besides the contribution of the soil moisture bridge, can the delayed effect of snow cover over the TP on summer rainfall be explained by the persistence of snow cover anomalies from the preceding seasons to summer?

Snow datasets used in the present study include the monthly satellite-derived snow water equivalent (SWE) dataset from November 1978 through May 2007 ( Armstrong et al., 2007), obtained from the National Snow and Ice Data Center (NSIDC). The data are gridded to the northern and southern 25 km Equal-Area Scalable Earth Grids (EASE-grids). With respect to those pixels with no microwave-derived SWE, the percent frequency of visible snow derived from weekly data is included in the dataset as a supplement. For convenience, the data were converted to regular 721 × 721 grids by setting the value at a regular grid to that at the nearest EASE-grid. We also used monthly station snow depth data from 174 stations in the TP within the territory of China, which were provided by the National Meteorological Information Center, China Meteorological Administration.

The precipitation dataset used in the present study is the monthly mean Climate Prediction Center (CPC) Merged Analysis of Precipitation ( CMAP; Xie and Arkin, 1997), which is available on 2.5° × 2.5° grids starting from January 1979. The SST dataset used in this study is the National Oceanic and Atmospheric Administration (NOAA) extended reconstructed version 3b monthly mean SSTs ( Smith et al., 2008).

Following Liu et al. (2014), the following process was carried out to calculate the snow cover extent based on the satellite-derived EASE-grid dataset. The snow cover grids with SWE over 1 mm and snow cover percentage over 35% were assigned the number one, and the grids with no snow cover were denoted using zero. The snow cover extent over the TP was derived from the synthetic information of the SWE and the snow cover percentage of visible snow in those pixels with no microwave-derived SWE. In this way, relevant information was retained to the maximum extent.

To quantitatively measure the variation of snow cover over the TP, an index was defined as SCAP of the TP region (25°N-43°N, 64°E-105°E), hereafter called the SCAP index ( Liu et al., 2014). In some research (e.g., Wu and Kirtman, 2007), the SWE is used without including the visible snow information in those pixels with no microwave-derived SWE. Therefore, the SWE averaged over the same TP region was considered as the SWE index and compared to the SCAP index. Moreover, an index defined as the station observational snow depth (SOSD) averaged over the TP region, called the SOSD index, was included for comparison. Based on these indices, the lag correlations between the preceding months (January-May) and the summer snow anomalies during the period 1979-2006 were calculated. As shown in Table 1, the SOSD index has a poor persistence with no significant correlations between the summer mean value and that in the preceding months. The SWE index exhibits a good persistence from January to summer with significant correlations. The SCAP index shows poor lag correlations from January to April, but the correlation is significant for May with a correlation coefficient of 0.63, indicating that the SCAP anomalies over the TP region can persist from May to summer.

Table 1

Table 1

Table 1 Lag correlation coefficients between the previous months and summer snow indices (the SCAP, SWE, and SOSD indices) for the period 1979-2006. The correlation coefficients at the 95% confidence level are in bold. Previous month SCAP index SWE index SOSD index January -0.09 0.52 0.13 February 0.27 0.46 0.06 March 0.32 0.43 0.27 April 0.34 0.52 0.01 May 0.63 0.74 0.33

Table 1 Lag correlation coefficients between the previous months and summer snow indices (the SCAP, SWE, and SOSD indices) for the period 1979-2006. The correlation coefficients at the 95% confidence level are in bold.

However, the high lag correlations in the SWE and SCAP indices might be due to the consistent decrease in snow in summer and the previous months, which is supported by the fact that the snow indices (the SWE and SCAP indices) show decreasing trends (data not shown). To reveal the persistence of the interannual variation of the snow indices, we removed the linear trends obtained through the least-squares fitting technique and further calculated the lag correlations (see Table 2). Similar to the original time series, no significant persistence appears in the SOSD index. Compared to the original time series, the lag correlations of the SWE index decrease remarkably, implying that the persistence of the SWE index comes primarily from the slow variation rather than from the interannual variation. As for the SCAP index, the lag correlations increase to the 95% confidence level in February through April. The lag correlation between the previous May and summer SCAP indices is up to 0.50 at the 99% confidence level. That is to say, the SCAP index shows a better persistence in the interannual variation, especially from the previous May to summer, relative to the other snow indices (the SWE and SOSD indices).

Table 2

Table 2

Table 2 Lag correlation coefficients between the previous months and summer snow indices (the SCAP, SWE, and SOSD indices) for the period 1979-2006, in which the linear trends of the indices were removed. The correlation coefficients at the 95% confidence level are in bold. Previous month SCAP index SWE index SOSD index January 0.17 0.25 0.12 February 0.63 0.39 0.03 March 0.37 0.28 0.18 April 0.39 0.39 -0.24 May 0.50 0.38 0.24

Table 2 Lag correlation coefficients between the previous months and summer snow indices (the SCAP, SWE, and SOSD indices) for the period 1979-2006, in which the linear trends of the indices were removed. The correlation coefficients at the 95% confidence level are in bold.

The May SCAP index displays a significant positive correlation with summer precipitation over the central region of eastern China and central Japan, i.e., the Meiyu- Baiu region (Fig. 2a), in good agreement with the correlation with the summer SCAP index (Fig. 1). Note that here the Meiyu region is slightly farther north than the classical Meiyu rain belt that is primarily located over the mid- and lower reaches of the Yangtze River Valley. On the basis of Fig. 1, summer precipitation averaged over the positive correlation regions, namely the Meiyu region (28°N-35°N, 107°E-120°E) and the Baiu region (35°N- 43°N, 128°E-154°E), is considered as an index to represent the variation in summer precipitation over the Meiyu-Baiu band, hereafter called the Meiyu-Baiu precipitation index. Correlation analysis shows that the summer Meiyu-Baiu precipitation index and the May SCAP index have a correlation coefficient of up to 0.51, which is at the 99% confidence level (Fig. 2b). These results support the theory that the May SCAP index has a close relationship with summer precipitation over the Meiyu-Baiu region and could be regarded as a forecasting factor. Certainly, the results strongly depend on the definition of the Meiyu-Baiu region. For example, since a different region is chosen to define the East Asian summer rainfall index ( Lu and Fu, 2010), the index does not show a good relationship with the May SCAP index, with a small correlation coefficient of 0.23 during the period 1979-2006.

Figure 2

	Figure Option ViewDownloadNew Window
	Figure 2 (a) Spatial distribution of correlation coefficients between the May SCAP index and summer precipitation for the period 1979-2006 with the shaded areas denoting correlation at the 90% confidence level. (b) The normalized time series of the May SCAP index (solid line with dots) and the summer Meiyu-Baiu precipitation index (dashed line with circles); correlation coefficient is 0.51.

It is well known that ENSO exerts a considerable influence on the following EASM ( Huang et al., 2004),which is also supported by the high correlation of winter SSTs in the tropical central-eastern Pacific with the summer precipitation over the Meiyu-Baiu region (not shown). Nevertheless, the May SCAP index has no significant relationship with, and is independent of, anomalous winter SSTs in the tropical central-eastern Pacific (not shown), implying that the close relationship between the SCAP and the precipitation cannot be simply attributed to their simultaneous responses to the ENSO. The summer SST anomalies in the tropical Indian Ocean may play an important role in modulating the vertical motion in situ and consequently affecting the western Pacific subtropical anticyclone (i.e., the western Pacific subtropical high (WPSH)) and associated EASM precipitation ( Xie et al., 2009; Liu et al., 2014). Moreover, the SST anomaly in the tropical Indian Ocean has a good persistence. Therefore, the forecasting ability of the preceding SST anomaly is worthy of investigation since the SST anomaly seems to affect the WPSH and associated summer precipitation over the Meiyu-Baiu region more directly than the SCAP index.

The correlation between the summer Meiyu-Baiu precipitation index and the May SSTs (Fig. 3a) displays a large significant positive area in the tropical Indian Ocean. This demonstrates that the May SST anomaly is an important precursory signal for summer Meiyu-Baiu precipitation. In turn, the May SST averaged over the tropical Indian Ocean (10°S-15°N, 40°E-100°E) has a significant positive correlation with summer precipitation over the Meiyu-Baiu region. This further verifies the close relationship between the May SST and summer precipitation anomalies (Fig. 3b). Nevertheless, a main body of significant positive correlation appears to the east of Japan (Fig. 3b), instead of covering central Japan as shown in Fig. 2a. That is to say, summer precipitation over the Baiu region of Japan seems to be better forecasted by using the May SCAP index than by using the May SST in the tropical Indian Ocean.

Figure 3

	Figure Option ViewDownloadNew Window
	Figure 3 Correlations (a) between the summer Meiyu-Baiu precipitation index and May SSTs and (b) between the May SST regionally averaged over the tropical Indian Ocean (the box region in (a)) and summer precipitation field. The shaded areas denote correlation at the 95% and 90% confidence level in (a) and (b), respectively.

One may wonder whether the May SCAP is independent of the May SST anomaly in the tropical Indian Ocean, which may to some degree reduce the importance of the SCAP index as a forecast factor. The correlation coefficient between the May SCAP index and the May SST in the tropical Indian Ocean is only 0.32, which is not significant. This indicates that the May SCAP over the TP is largely independent of the tropical Indian Ocean SST.

To further reveal the independent effect of the May SCAP on the summer precipitation over East Asia, we examine a partial correlation of summer precipitation with the May SCAP index after removing the effects of the May SST anomaly in the tropical Indian Ocean SST (Fig. 4). The partial correlation was calculated using the following formula ( Zar, 1998; Wu and Kirtman, 2007): ,where r_ij refers to the correlation between i and j; and indices 1, 2, and 3 represent summer precipitation, the May SCAP index, and the May SST regionally averaged over the tropical Indian Ocean, respectively. Figure 4 shows that significant correlations appear over central China and Japan, which generally resembles Fig. 2a. Although the significant correlation areas are slightly smaller than those in Fig. 2a, the result indicates that the effect of the May SCAP is still strong enough to modulate summer precipitation over the Meiyu-Baiu region without the effect of the May SST anomaly in the tropical Indian Ocean.

Figure 4

	Figure Option ViewDownloadNew Window
	Figure 4 Partial correlation of summer precipitation with the May SCAP index after removing the effects of the May SST anomaly in the tropical Indian Ocean. The shaded areas denote correlation significant at the 90% confidence level.

In this study, we investigated the persistence of the snow anomaly over the TP from the preceding seasons to summer in different datasets, and explored the relationship between the previous snow anomaly and summer precipitation over East Asia. The results showed that SCAP index has a higher persistence in interannual anomalies, particularly from May to summer, than the other snow indices, such as the SOSD index and the SWE index. Because of its high persistence, the SCAP index in May is significantly related to summer precipitation over the Meiyu-Baiu region and could be regarded as a forecast factor. The persistence of the SCAP index can partly explain the season-delayed effect of snow cover over the TP on summer rainfall over the Meiyu-Baiu region besides the contribution of the soil moisture bridge.

Certainly, the previous SST anomaly in the tropical Indian Ocean can persist through summer and affect summer precipitation over the Meiyu-Baiu region. However, the May SCAP index is generally independent of the simultaneous SSTs in the tropical Indian Ocean. Partial correlation analysis suggested that the effect of the May SCAP may affect summer precipitation over the Meiyu- Baiu region independent of the effects of the SST anomaly in the tropical Indian Ocean. Therefore, the forecast for summer precipitation over the Meiyu-Baiu region by synthetically using the preceding SST signal and the preceding SCAP signal over the TP is worthy of further exploration.