A Diagnostic Study of the Influence of Early Spring Soil Moisture in Southeastern China on Interannual Variability of the East Asian Subtropical Summer Monsoon Onset

As one component of the Asian monsoon system, the East Asian monsoon has a highly complex temporal and spatial structure (Tao and Chen, 1987). The East Asian summer monsoon contains the East Asian tropical summer monsoon (i.e., the South China Sea summer monsoon) and the East Asian subtropical summer monsoon (EASSM) (Zhu et al., 1986; Wang and LinHo, 2002). Different East Asian monsoon subcomponents interact with each other and often result in droughts and floods over East Asia (Zhang and Tao, 1998).

The onset of the EASSM is accompanied by the appearance of southwesterly winds in the East Asian subtropical region. It is also marked by the seasonal transi-tion in the zonal thermal contrast between the East Asian continent and the western North Pacific Ocean to the south of the Yangtze River, and the emergence of convective precipitation over the southeastern China around late March and early April (He et al., 2007; Zhao et al., 2007; Qi et al., 2008). Zhu et al. (2012) pointed out that the formation of the EASSM is induced by the seasonal change in zonal sea–land thermal contrast in the subtro-pics. As the land surface sensible heating enhances ra-pidly, temperature in the low-level atmosphere over East Asia shows a zonal structure of “warm in the west and cold in the east.” Such a pattern strengthens the low-level southerly winds prevailing over the southern China. The low-level southerly winds enhance the vertical northerly shear, resulting in ascending motion and rainfall locally. The latent heat of condensation released from the rainfall can further strengthen the low-level southerly winds. Such a positive feedback finally leads to the formation of the EASSM.

As one of the most fundamental parameters in land surface processes, soil moisture (SM) has a substantial influence on climate, and its importance has been widely recognized (Dirmeyer et al., 2009). SM can affect the surface energy and water over East Asia through altering latent heat, sensible heat, surface albedo, thermal capacity, plant growth, and evapotranspiration, which has the potential to change local climate (Ma et al., 2000, 2001; Seneviratne et al., 2010; Zhang and Zuo, 2011; Li and Ma, 2013; Liu et al., 2016; Chen et al., 2019, 2020; Dong et al., 2022). The variability of summer climate over East Asia is closely related with the spring SM anomalies over East Asia (Zhang and Zuo, 2011; Liu et al., 2017) and Eurasian continent (Zhang et al., 2017). For example, Zuo and Zhang (2007) and Zhang and Zuo (2011) pointed out that when soil in the region from the lower and middle reaches of the Yangtze River valley to North China is wet in spring, the local energy balance changes and leads to the decrease of surface temperature at the end of spring, which weakens the East Asian summer monsoon. Such physical process was also verified in numerical simulations by Zuo and Zhang (2016). The local SM anomalies in the southern and southwestern China in summer may cause strong land–atmosphere interaction and play an important role in the variability of the summer precipitation and circulation over East Asia (Deng et al., 2018; Gao et al., 2018; Zeng and Yuan, 2018; Dong et al., 2022). Some studies have shown that spring SM anomalies in the Indo–China Peninsula can modulate the onset and development of South China Sea summer monsoon and the atmospheric circulation in the surrounding areas (Zhang and Qian, 2002; Sen et al., 2004; Wu et al., 2014; Pan et al., 2017; Ma et al., 2018; Gao et al., 2019, 2020).

As mentioned above, SM anomalies play an important role in the variability of East Asian summer monsoon. However, there are no studies concerning the influence of SM on the EASSM onset until now. Previous studies stressed that the onset of EASSM is mainly caused by the seasonal variation of the subtropical zonal thermal difference between land and sea, but the role played by SM in the land–sea thermal contrast during the EASSM onset has not been investigated before. Because the thermal condition of land can be affected by SM, we propose the hypothesis that the SM anomalies over land may have important impact on the sea–land thermal contrast and thus affect the EASSM onset.

This work thus intends to investigate the influence of SM on the onset of EASSM, which will be of great scientific significance in understanding the mechanism of monsoon variability and also has the value in practical application for monsoon prediction. The rest of this study is organized as follows. Section 2 describes the datasets and analysis methods used in this study. We explore and define the criteria for the onset date of EASSM in Section 3. Linkages between SM anomalies over the land region in subtropical East Asia and the onset date of EASSM are examined in Section 4. In Section 5, we propose a possible physical mechanism that the SM anomalies affects the onset date of EASSM. Finally, conclusions and discussion are presented in Section 6.

Due to the shortage of daily soil moisture observational data in China, an alternative way is to use reanalysis data to make up for it. According to the study of Liu et al. (2014), SM data from European Center for the ECMWF reanalysis (ERA-Interim; Dee et al., 2011) can well reproduce the observed temporal and spatial variation characteristics of SM in the eastern China. Considering the uncertainty of SM data, in this study the SM data from the ECMWF reanalysis v5 (ERA5; Hersbach et al., 2020), the NOAH model data of the Global Land Data Assimilation System (GLDAS-NOAH; Ek et al., 2003), and the NCEP Climate Forecast System Reanalysis (CFSR; Saha et al., 2010) are also employed. The present study uses SM data of ERA-Interim and ERA5 at a depth of 0–7 cm with a horizontal resolution of 0.5° × 0.5°, and those of CFSR and GLDAS-NOAH at a depth of 0–10 cm with horizontal resolutions of 0.5° × 0.5° and 1.0° × 1.0°, respectively. By using the water density of 10³ kg m⁻³, the unit of all data is unified as m³ m⁻³. Considering the availability of SM data, we choose 1981–2010 as the research period. In addition, to assess the reliability of the surface flux (sensible heat, latent heat, radiative fluxes, and skin temperature), we utilized three sets of the daily reanalysis data from the 1) ERA-Interim with a horizontal resolution of 0.5° × 0.5°, 2) GLDAS-NOAH with a horizontal resolution of 1.0° × 1.0°, and 3) NCEP/NCAR with a horizontal resolution of 1.875° × 1.875° (Kalnay et al., 1996). We also use precipitation data from pentad-scale Climate Prediction Center Merged Analysis of Precipitation (CMAP) (Xie and Arkin, 1997), and temperature and wind fields from NCEP/NCAR reanalysis with 17 vertical pressure levels from 1000 to 10 hPa.

In this study, we aggregate various daily reanalysis data to pentads by averaging the data in every 5 days as 1 pentad. The 12th pentad in leap years is defined as a 6-day average from 25 February to 1 March. The main statistical methods utilized in this study are commonly used ones including Pearson correlation analysis, linear regression, composite analysis, and a two tailed Student’s t-test. In addition, for descriptive convenience, we define the onset pentad of EASSM as 0 pentad, and the 1-pentad leading and lagging the monsoon onset pentad as −1 and +1 pentads, respectively, and so on.

The surface energy balance is calculated by the following formula (Lu and Cai, 2009):

where Q is the heat storage term at land surface, $ {S}^{\uparrow } $ and $ {S}^{\downarrow } $ represent surface upward and downward shortwave radiations, respectively, $ {F}^{\uparrow } $ and $ {F}^{\downarrow } $ represent surface upward and downward longwave radiations, respectively, and H and LE are surface sensible and latent heat fluxes, respectively.

Figure 1 shows latitude–time cross-section of the climatological rainfall and 850-hPa winds averaged over 110°–125°E, and the time series of climatological 850-hPa meridional winds averaged over the subtropical region 20°–30°N, 110°–125°E. It can be seen in Fig. 1 that the precipitation in the subtropical region of East Asia increases continuously from the beginning of the year along with the enhancing of southerly winds. When the southerly winds reach about 2 m s⁻¹ in the 14th pentad, a marked increase of precipitation to more than 5 mm day⁻¹ occurs. The strong precipitation is maintained from the 14th to 32nd pentads, with the southerly winds remaining at about 2 m s⁻¹. After the 32nd pentad, the southerly winds increase sharply and the strong precipitation moves to the Yangtze River valley, indicating the beginning of Meiyu season in China. Li and Zhang (2012) proposed that the increase of precipitation in the subtropical region of the eastern China from January to about mid-March is caused by the strengthening of bypass flow around the Tibetan Plateau, and from around mid-March the increased precipitation is associated with the onset of EASSM. Therefore, the southerly winds maintaining stably at about 2 m s⁻¹ over the subtropical region of East Asia appear to be a distinguishing characteristic of the EASSM.

Figure 1.  Latitude–time cross-section of climatological precipitation (shadings; mm day⁻¹) and 850-hPa winds (vectors; m s⁻¹) along 110°–125°E, and the time series of the climatological 850-hPa meridional winds (red line; m s⁻¹) averaged over the subtropical region 20°–30°N, 110°–125°E during 1981–2010. The latitude and meridional wind speed are marked in the y-axes on the right and left, respectively. The heavy dashed black line and white line are 0 and 2 m s⁻², respectively, for the climatological 850-hPa meridional winds averaged over the subtropical region 20°–30°N, 110°–125°E.

In light of the seasonal evolution of the meridional winds at 850 hPa shown in Fig. 1, the onset date of EASSM can be defined by the first date after the 13th pentad when the average meridional southerly wind speed exceeds 2 m s⁻¹ in the region of 20°–30°N, 110°–125°E after lasting at least 2 pentads. Based on this definition, the time series of EASSM onset date during 1981–2010 is obtained and shown in Fig. 2. It can be seen that the EASSM onset date has a remarkable interannual variation, and the average onset date of EASSM is pentad 17.7 during 1981–2010, about 26 March.

Figure 2.  Time series of EASSM onset date during 1981–2010. The heavy dashed line is the climatological EASSM onset date at the 17.7 pentad.

Although some previous studies have defined the onset date of the EASSM by the vertical meridional wind shear (Zhu et al., 2012; Chi et al., 2015) and by the soaking rainfall (a continuous 20 mm precipitation) (Huang et al., 2009; Tang and Li, 2020), our definition contains the features of both the lower tropospheric meridional winds and precipitation over the East Asia subtropical region associated with the onset of the EASSM. The onset date defined by us is well correlated with the EASSM onset date defined by Huang et al. (2009) using the soaking rainfall, with a correlation coefficient being 0.65, exceeding the 0.1 significance level determined by a two tailed Student’s t-test.

In order to explore the relationship between SM and EASSM onset date, the correlation coefficients between the EASSM onset date and the preceding SM in the eastern China in about 1 month prior to the onset date (averaged from preceding −6 to −1 pentad) during 1981–2010 are calculated by using ERA-Interim, ERA5, GLDAS-NOAH, and CFSR SM data, respectively (Figs. 3a–d). It can be seen that the correlation coefficients between the preceding SM and EASSM onset date in different datasets are markedly different in the region north of about 30°N, indicating that there may be a great uncertainty in SM data in these areas. However, in the southeastern China to the south of about 30°N, the correlation coefficients in the four sets of reanalysis data are obviously consistent. In each dataset, there is a noteworthy positive correlation between the preceding SM and EASSM onset date, that is, when the SM in the month preceding onset is abnormally high/low in the southeastern China, the EASSM onset date is abnormally delayed/advanced.

Figure 3.  Spatial distributions of correlation coefficients between EASSM onset date and SM in the eastern China in the leading time averaged in pentads (a–d) −6 to −1, (e–h) −4 to −1, and (i–l) −8 to −1. SM data are from (a, e, i) ERA-Interim, (b, f, j) ERA5, (c, g, k) GLDAS-NOAH, and (d, h, l) CFSR data. Blue box in the upper panel denotes the key region of SM over 22.5°–27.5°N, 110°–117.5°E. Black dots indicate the correlation coefficients exceeding the 0.1 significance level determined by a two-tailed Student’s t-test.

To further assess whether the significant positive correlation in the southeastern China is affected by the leading time of SM, the correlation coefficients between the EASSM onset date and the SM in the eastern China with a leading time of 2 pentads less than 1 month (averaged from −4 to −1 pentad) and 2 pentads more than 1 month (from −8 to −1 pentad) during 1981–2020 are calculated, and the results are given in Figs. 3e–l, respectively. To compare the spatial distributions of the correlation coefficients between the EASSM onset date and the SM with different lengths of leading time, the results in the same dataset have a highly consistent behavior, indicating that the correlation of the EASSM onset date with the SM in the preceding month is not sensitive to the length of the preceding time. Therefore, in the following we will investigate the relations of EASSM onset with the SM leading 1 month (averaged from −6 to −1 pentad) in the southeastern China.

To further illustrate the consistency in the interannual variations of different SM data in the southeastern China, we select the key region 22.5°–27.5°N, 110°–117.5°E (the blue box in Figs. 3a–d) and exhibit the time series of the standardized SM anomalies averaged from −6 to −1 pentad over the key region (Fig. 4). The results show that the four datasets have a high agreement on the variation in the interannual timescale, the correlation coefficients between any two data are higher than 0.89, exceeding the 0.01 significance level determined by two-tailed Student’s t-tests. Therefore, the ERA-Interim SM dataset is used to select five high (low) SM years with standard deviations higher than 0.8 (lower than −0.8). The years of 1983, 1985, 1992, 1995, and 1998 are rated as the SM high years, and 1987, 1996, 1999, 2002, and 2008 as the SM low years.

Figure 4.  Time series of the standardized SM anomalies averaged from −6 to −1 pentad over the key region 22.5°–27.5°N, 110°–117.5°E from ERA-Interim (red), ERA5 (yellow), GLDAS-NOAH (blue), and CFSR (green) data during 1981–2010. The upper and lower heavy dashed lines indicate the standard deviation equal to 0.8 and −0.8, respectively.

As shown in the above analyses, the onset date of EASSM has a remarkable positive correlation with the preceding SM averaged from −6 to −1 pentad over the southeastern China. To show the persistence of SM anomalies from −6 to −1 pentad, in Fig. 5 we depict the spatial distributions of correlation coefficients of the time series of the SM anomalies averaged from −6 to −4 pentad over the key region with SM anomalies averaged from −6 to −4 pentad (Fig. 5a), from −4 to −2 pentad (Fig. 5b), and from −2 to 0 pentad (Fig. 5c) over the eastern China, respectively. It is evident that significant positive correlations occur over the southeastern China from −6 to −4 pentad and from −2 to 0 pentad. It is demonstrated that the abnormal signals of SM anomalies can be continuously transmitted to the 0 pentad when the EASSM onset occurs because of the “memory” characteristics of SM.

Figure 5.  Spatial distributions of correlation coefficients between the time series of SM averaged from −6 to −4 pentad over the key region 22.5°–27.5°N, 110°–117.5°E and the SM averaged (a) from −6 to −4 pentad, (b) from −4 to −2 pentad, and (c) from −2 to 0 pentad over the eastern China. Blue boxes denote the key region. Black dots indicate the correlation coefficients exceeding the 0.01 significance level determined by a two-tailed Student’s t-test.

As indicated in the above analyses, the onset date of EASSM has a significant positive correlation with the SM over the southeastern China in the month preceding onset. Positive (negative) preceding SM anomalies in the southeastern China are associated closely with the delayed (advanced) onset of EASSM. In this section, we will reveal how the preceding SM over the southeastern China affects the onset date of EASSM. Considering SM is an important parameter in affecting the surface thermal process, the land thermal condition associated with the SM anomalies over the southeastern China will be investigated.

The variation of SM can affect surface thermal condition by altering the exchanges of energy and water between soil and atmosphere. To further substantiate this influence over the southeastern China, Fig. 6 shows the composite differences between the preceding SM high and low years in the month preceding onset (averaged from −6 to −1 pentad) for SM (Fig. 6a), net upward shortwave radiation (Fig. 6b), net upward longwave radiation (Fig. 6c), total cloud cover (Fig. 6d), sensible heat flux (Fig. 6e), latent heat flux (Fig. 6f), heat storage in the surface (Fig. 6g), and surface temperature (Fig. 6h). The surface fluxes come from ERA-Interim dataset, in which the climatological mean on each pentad from 1981 to 2010 is removed. It is clear that the significant positive SM difference can be observed in the southeastern China (Fig. 6a), which changes the surface energy balance. When the SM is abnormally high in the southeastern China before the onset of EASSM, the upward shortwave radiation strengthens (Fig. 6b). Meanwhile, the upward longwave radiation (Fig. 6c), sensible heat (Fig. 6e), and latent heat (Fig. 6f) decrease. In addition, to further investigate the surface energy budget, the heat storage term (Q) is calculated according to Eq. (1) and the result is shown in Fig. 6g. The variable Q is positive in the southeastern China, representing the upward transport of total surface heat, indicating that the abnormally high (low) SM is associated with an abnormally low (high) surface temperature in the southeastern China (Fig. 6h). The strengthened upward shortwave radiation appeared in higher SM years (Fig. 6b) may be attributed to the increase of clouds (Fig. 6d). In higher SM years, the increased clouds may reduce the absorbed shortwave radiation in the surface, which prevents the heating of the surface from the shortwave radiation and is favorable for a reduction of surface temperature. In most cases, change in sensible heat is usually opposite to that of latent heat. However, sensible heat and latent heat are positively correlated in southeastern China from −6 to −1 pentad (figure omitted). Therefore, sensible heat and latent heat decrease in SM high years.

Figure 6.  Composite differences between SM high and low years averaged from −6 to −1 pentad for (a) SM (m³ m⁻³), (b) net upward shortwave radiation (SR; W m⁻²), (c) net upward longwave radiation (LR; W m⁻²), (d) total cloud cover (TCC; %), (e) sensible heat flux (SH; W m⁻²), (f) latent heat flux (LH; W m⁻²), (g) heat storage at land surface (Q; W m⁻²), and (h) surface temperature (ST; K). Black box denotes the key region of SM. Dots indicate the composite differences exceeding the 0.1 significance level. Surface flux is from ERA-Interim dataset after removing the climatological mean on each pentad from 1981 to 2010.

The SM-related heat flux and Q are examined quantitatively based on area-mean values over the key region (Fig. 7). It is apparent that net shortwave radiation dominates in the southeastern China exceeds 30 W m⁻². The contributions of longwave radiation, sensible heat, and latent heat are all negative and almost consistent. The sum of the four terms obtains Q, which is a positive value. This result is consistent with that in Fig. 6. When the SM is abnormally high (low) in the southeastern China before the onset of EASSM, the local heat flux is transported upward (downward), leading to a decrease (increase) in surface temperature.

Figure 7.  Composite differences between SM high and low years in net upward shortwave radiation (SR; W m⁻²), net upward longwave radiation (LR; W m⁻²), sensible heat flux (SH; W m⁻²), latent heat flux (LH; W m⁻²), and heat storage (Q; W m⁻²) averaged from −6 to −1 pentad over the key region.

Figure 8 shows the scatter diagram of standardized SM anomalies, surface fluxes, heat storage, and surface temperature anomalies averaged from −6 to −1 pentad in the key region in the southeastern China during 1981–2010. The SM, surface fluxes, and surface temperature come from ERA-Interim dataset. It is obvious that the preceding SM over the southeastern China has a significant positive correlation with net upward shortwave radiation (Fig. 8a) and heat storage (Fig. 8e), and significant negative correlations with net upward longwave radiation (Fig. 8b) and sensible heat flux (Fig. 8c). All correlation coefficients exceed the 0.01 significance level. Furthermore, the preceding SM (ERA-Interim) in the southeastern China shows a significant negative correlation with surface air temperature in the southeastern China (Fig. 8f), indicating that the abnormally wet (dry) soil reduces (increases) surface air temperature. These results on the interannual timescale further prove the results of the composite analysis shown in Fig. 6, demonstrating that the energy exchange between land and air in the southeastern China is sensitive to SM anomalies.

Figure 8.  Scatter diagrams of standardized SM anomalies and (a) net upward shortwave radiation (SR), (b) net upward longwave radiation (LR), (c) sensible heat flux (SH), (d) latent heat flux (LH), (e) heat storage (Q), and (f) surface temperature (ST) averaged from −6 to −1 pentad in the key region in the southeastern China during 1981–2010. SM and surface flux are both from ERA-Interim data after removing the climatological mean on each pentad from 1981 to 2010. Red lines represent the regressions for ERA-Interim. * and *** indicate that the correlation coefficients exceed the significance level of 0.1 and 0.01, respectively.

However, as seen in Fig. 8d, the preceding SM is not well correlated with latent heat flux, and the correlation coefficient between latent heat and SM is only 0.23. Such phenomenon indicates that a decreased (increased) latent heat flux corresponding to a wetter (drier) soil in Fig. 6d may be uncertain. To verify the reliability of surface fluxes and surface temperature in reanalysis data, the surface fluxes and surface temperature data from GLDAS-NOAH and NCEP/NCAR are further compared. As shown in Fig. 9, it is clear that the shortwave radiation, longwave radiation, sensible heat flux, and surface temperature show good consistencies on the interannual time scale with statistically significant correlations between any two of them exceeding the 0.05 significance level except the correlation coefficient for sensible heat flux between GLDAS-NOAH and NCEP/NCAR (Fig. 9c). Nevertheless, for the latent heat fluxes (Fig. 9d), ERA-Interim and GLDAS-NOAH latent heat fluxes have significant correlation with the correlation coefficient being 0.77, exceeding the 0.05 significance level. But the correlation coefficients of the latent heat flux of the NCEP/NCAR with those of ERA-Interim and GLDAS-NOAH are not significant (0.12 and 0.25, respectively). Therefore, the poor correlations between the preceding SM and latent heat flux shown in Fig. 7d may be attributed to the data quality of latent heat.

Figure 9.  Time series of standardized (a) net surface shortwave radiation, (b) net surface longwave radiation, (c) sensible heat flux, (d) latent heat flux, and (e) surface temperature averaged from −6 to −1 pentad over the key region in the southeastern China from ERA-Interim (red), GLDAS-NOAH (blue), and NCEP/NCAR (green) during 1981–2010. The R(E,G), R(E,N), and R(G,N) in the figure represent the correlation coefficients between the data of ERA-Interim and GLDAS-NOAH, ERA-Interim and NCEP/NCAR, and GLDAS-NOAH and NCEP/NCAR, respectively. ** indicates that the correlation coefficient exceed the significance level of 0.05.

The above analyses show that the abnormally wet (dry) soil in the southeastern China in the month preceding onset is associated with the decrease (increase) of surface temperature through changing the local surface energy balance. Considering that the reversal of zonal sea–land temperature contrast in the East Asian subtropical region is the fundamental cause for the onset of EASSM (Zhao et al., 2007; He et al., 2008), we calculate the sea–land surface temperature contrast (sea surface temperature minus land surface temperature) between the ocean region of 22.5°–27.5°N, 122.5°–130°E in the western North Pacific and the land region of 22.5°–27.5°N, 110°–117.5°E in the southeastern China. The time series of correlation coefficients between 850-hPa meridional winds averaged over the subtropical region and the sea–land temperature differences are given in Fig. 10. From mid-February to mid-May, the negative correlation coefficients between the meridional winds and the sea–land temperature difference exceed the 0.1 significance level. A larger (smaller) sea–land temperature contrast corresponds to a weaker (stronger) meridional southerly winds over the subtropical region from mid-February to mid-May, indicating a delayed (advanced) EASSM onset.

Figure 10.  Time series of correlation coefficients between the 850-hPa meridional winds over the subtropical region (20°–30°N, 110°–125°E) and the sea–land temperature difference from the 6th to 30th pentad during 1981–2010. The sea–land temperature difference is the sea surface temperature averaged in 22.5°–27.5°N, 122.5°–130°E minus the land surface temperature averaged in 22.5°–27.5°N, 110°–117.5°E. The three red dashed lines indicate the correlation coefficients equal to the 0.01, 0.05, and 0.1 significance levels, respectively, determined by a two-tailed Student’s t-test.

Figure 11 shows the time series of composite differences between SM high and low years for sea–land temperature contrast. It should be noted that the sea–land temperature contrast in SM high years was significantly higher than SM low years during the 8th–18th pentads, and such difference in sea–land temperature difference can lead to the interannual difference of the onset date of EASSM. As shown in Fig. 3, the high (low) preceding SM in the southeastern China correspond to a later (earlier) onset of EASSM. The abnormally wet (dry) soil over the southeastern China results in a lower (higher) local surface temperature and reduced (increased) the sea–land temperature contrast, which delays (advances) the seasonal transition of meridional southerly winds around the East Asian coast, and is conducive to a later (earlier) onset of EASSM.

Figure 11.  Time series of composite difference between SM high and low years for the sea–land temperature difference (K) from the 1st to 36th pentad. The right and left blue heavy lines represent the 8th and 18th pentad, respectively.

Considering that the increases (decreases) of surface temperature can transfer the heat from the surface to the atmosphere and enhances (reduces) the heat received by the low-level atmosphere, in Fig. 12 we give the longitude–pentad cross-section of the vertically integrated air temperature from 1000 to 500 hPa averaged over 22.5°–27.5°N in the southeastern China in SM high years (Fig. 12a) and SM low years (Fig. 12b) as well as their difference (Fig. 12c). It is obvious that the negative and positive anomalies of the low-level air temperature in the subtropical region of East Asia (100°–120°E) appear before the onset of EASSM in about the 18th pentad in SM high and low years, respectively, and the negative difference between them is statistically significant. In fact, a strengthened (weakened) sea–land temperature contrast corresponds to a reduced (enhanced) southerly winds at 850 hPa over the subtropical region from mid-February to mid-May (Fig. 10), and the sea–land temperature contrast in SM high years was significantly higher than SM low years (Fig. 11). Therefore, an abnormally wet (dry) soil reduces (increases) the warm and moist air coming from the south in the lower troposphere and result in a cold (warm) air temperature in the low-level atmosphere.

Figure 12.  Longitude–time cross-section of vertically integrated air temperature from 1000 to 500 hPa averaged in 22.5°–27.5°N for (a) high and (b) low SM years as well as (c) their difference (K). Dots in (c) indicate the composite differences exceeding the 0.1 significance level.

The low air temperature in the low-level atmosphere over the southeastern China corresponding to high SM (Fig. 12a) is not conducive to the enhancement of southerly winds and results in a delayed onset of EASSM, while in the SM low years (Fig. 12b), the high air temperature in the low-level atmosphere is conducive to an advanced onset of EASSM. The sea–land thermal difference in the low-level atmosphere is similar to the thermal difference at the surface (figure not shown), which further verified the previous conclusion that the preceding SM anomalies in the southeastern China has significant impact on the onset of EASSM through affecting the sea–land thermal difference.

In Fig. 13, the regressed 850-hPa winds against the sea–land temperature contrast in March are shown. As seen in this figure, a strengthened sea–land temperature contrast is well related with the anomalous northerly winds over the southeastern China, indicating that a wet SM is conducive to a delayed onset of EASSM. The appearance of anomalous northerly winds over the wet SM is physically understandable. In the subtropics, the vorticity equation can be simplified and expressed as: $ \beta v\propto (f+\mathrm{\zeta })/{\theta }_{z}\cdot \partial Q/\partial z $ (Liu et al., 1999, 2001; Wei et al., 2014), where v is the meridional winds, f and β are the Coriolis parameter and its meridional gradient respectively, ζ is the vertical vorticity, Q is the diabatic heating, and θ_z is the static stability. Because of the anomalous cooling at the surface of the wet SM (Fig. 6h), the $ \partial Q/\partial z < 0 $, and thus v < 0. Therefore, the anomalous northerly winds prevail over the southeastern China.

Figure 13.  Regressed 850-hPa winds (vectors; m s⁻¹) against the sea–land temperature contrast in March during 1981–2010. The sea–land temperature contrast is the sea surface temperature averaged in 22.5°–27.5°N, 122.5°–130°E minus the land surface temperature averaged in 22.5°–27.5°N, 110°–117.5°E. The shaded areas indicate regression coefficients exceeding the 0.1 significance level.

Through the above analysis, the physical explanation can be given to the later EASSM onset date in the SM high years than in the SM low years. The abnormally wet (dry) soil over the southeastern China can lead to a lower (higher) local surface air temperature, which weakens (strengthens) the zonal sea–land thermal difference in surface and low-level atmosphere. The increased (decreased) sea–land thermal difference is favorable for anomalous northerly (southerly) winds in the lower troposphere in the wet (dry) SM over the southeastern China, resulting in a delayed (advanced) onset of EASSM.

In this study, we define a criterion for the onset date of EASSM according to the characteristics of the 850-hPa meridional winds over the East Asian subtropical region that a 2-m s⁻¹ southerly winds maintain in spring and last for about 3 months during the period of 1981–2010. The defined onset date of EASSM has obvious interannual variation and the average onset date is in pentad 17.7. Based on four sets of SM reanalysis data, the correlation coefficients between the EASSM onset date and preceding SM in the eastern China are calculated. It is found that the EASSM onset date is significantly correlated with the SM in the southeastern China in the month preceding onset. Positive (negative) preceding SM anomalies in the southeastern China are conducive to a delayed (advanced) onset of EASSM. In addition, the SM anomalies could well persist from pentad −6 before the onset of EASSM to pentad 0 when the onset occurs.

The possible physical mechanism of the preceding SM anomalies in the southeast China affecting the onset date of EASSM is also explored. It is found that the abnormally wet (dry) soil alters the surface energy balance by adjusting the total surface heat flux, accompanying with the decreasing (increasing) of surface temperature. The decreased (increased) surface temperature in the southeastern China weakens (increases) the zonal sea–land thermal contrast in the surface and low-level atmosphere, resulting in a later (earlier) EASSM onset.

Previous studies demonstrated that the zonal sea–land temperature difference in the East Asian subtropical region is the fundamental reason for the climatological onset of EASSM (Zhao et al., 2007; He et al., 2008). This study indicates that the zonal sea–land temperature difference is affected by the SM anomalies in the southeastern China, and thus impacts on the interannual variability of the EASSM onset. This study not only gives new idea in understanding the EASSM onset, but also provides a new and useful idea for predicting the onset of EASSM by the SM anomalies in the southeastern China.