HTML

The 0.5° × 0.5° gridded observations of daily precipitation covering global land areas developed by the US Climate Prediction Center (CPC) are used in this study. This gridded precipitation product came from rain gauge observations over global land areas (Chen et al., 2008). Daily mean outgoing longwave radiation (OLR) on a 2.5° × 2.5° spatial resolution from the National Oceanic and Atmospheric Administration (NOAA) is used (Liebmann and Smith, 1996). The daily atmospheric elements are from the ECMWF interim reanalysis (ERAInterim; Dee et al., 2011) at a resolution of 0.75° × 0.75°. All datasets cover the period 1979–2016.

The diabatic heating
${Q_1}$ is calculated according to the method introduced by Yanai et al. (1973) and Yanai and Tomita (1998):$${Q_1} = \frac{{\partial T}}{{\partial t}} + {{{V}}_{\rm{h}}} \cdot {\nabla _{\rm{h}}}T + \omega \left({\frac{{\partial T}}{{\partial p}}  \frac{{RT}}{{{c_p}p}}} \right), $$ (1) where
$T$ ,$R\;\;$ , and${c_p}$ are the air temperature, the gas constant for dry air, and the specific heat of dry air at constant pressure, respectively.${{{V}}_{\rm{h}}} = (u, v)$ is the horizontal wind,$\omega $ is the vertical velocity in pressure (p) coordinates, and${\nabla _{\rm{h}}}$ is the horizontal gradient operator. Vertical integration of${Q_1}$ is from the tropopause pressure (${p_{\rm {t}}}$ ) to the surface pressure (${p_{\rm {s}}}$ ) as follows,$$\left\langle {\left. {{Q_1}} \right\rangle } \right. = \frac{{{c_p}}}{g}\int_{{p_{\rm {t}}}}^{{p_{\rm {s}}}} {Q{}_1{\rm{d}}p}, $$ (2) where
$g$ is the gravitational acceleration and${p_{\rm {t}}}$ is set to be 100 hPa.The anticyclonic circulation corresponds to low PV values. In terms of isobaric PV, if ignoring the friction term, the local tendency of PV is expressed as (Zhang and Ling, 2012; Ren et al., 2015)
$$\frac{{\partial {\rm{PV}}}}{{\partial t}} =  \underbrace {\left( {u\frac{{\partial {\rm PV}}}{{\partial x}} + v\frac{{\partial \rm{PV}}}{{\partial y}}} \right)}_{{\rm{part}} {\text{}} h}  \underbrace {\omega \frac{{\partial \rm{PV}}}{{\partial p}}}_{{\rm{part}} {\text{}} w}  \underbrace {g\left( {f + \zeta } \right)\frac{{\partial {Q_1}}}{{\partial p}}}_{{\rm{PVG}}}  g\left( {\zeta _x^{}\frac{{\partial {Q_1}}}{{\partial x}} + \zeta _y^{}\frac{{\partial {Q_1}}}{{\partial y}}} \right), $$ (3) where
${\zeta _{{x}}}$ is the zonal relative vorticity and${\zeta _y}$ the meridional one. The term$  (u\displaystyle\frac{{\partial \rm{PV}}}{{\partial x}} + v\displaystyle\frac{{\partial \rm{PV}}}{{\partial y}})$ is PV advection via horizontal winds (parth). The term$  (\omega \displaystyle\frac{{\partial \rm{PV}}}{{\partial p}})$ is PV vertical advection (partw). The term$  g(f + \zeta)\displaystyle\frac{{\partial {Q_1}}}{{\partial p}}$ is nonuniform distribution of${Q_1}$ in the vertical direction. It is a PV generation (PVG) term. The last term in Eq. (3) is also a PV generation term via horizontal nonuniform distribution of${Q_1}$ . The last term is not considered due to its much smaller value than the others. Thus, PV generation results mainly from the vertical gradient of${Q_1}$ .By using the method introduced by Zhang and Ling (2012) and Ren et al. (2015), a variable is divided into its time mean (denoted by an overbar) and subseasonal (denoted by a prime) and nonsubseasonal components. The subseasonal anomalies of local PV changes (
$ {\partial {\rm PV}'}/{\partial t}$ ) are decided by the following processes:$$\!\!\!\!\begin{split} {\left( {\displaystyle\frac{{\partial {\rm PV}'}}{{\partial t}}} \right)_{{\rm{part}} {\text{}} h}} = &  \left( {\bar u\displaystyle\frac{{\partial {\rm PV}'}}{{\partial x}} + \bar v\displaystyle\frac{{\partial {\rm PV}'}}{{\partial y}}} \right)  \\ & \left( {u'\displaystyle\frac{{\partial \overline {\rm PV} }}{{\partial x}} + v'\displaystyle\frac{{\partial \overline {\rm PV} }}{{\partial y}}} \right) + {\rm{residue}}{\rm{ {\text{}} }}h,\quad\quad\quad \end{split}$$ (4) $$\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!{\left({\frac{{\partial {\rm PV}'}}{{\partial t}}} \right)_{{\rm{part}} {\text{}} w}} =  \overline \omega \frac{{\partial {\rm PV}'}}{{\partial p}}  \omega '\frac{{\partial \overline {\rm PV} }}{{\partial p}} + {\rm{residue}}{{ {\text{}} w}}, $$ (5) $${\left({\frac{{\partial {\rm PV}'}}{{\partial t}}} \right)_{{\rm{PVG}}}} =  g(f + \overline \zeta)\frac{{\partial {Q_1}'}}{{\partial p}}  g\zeta '\frac{{\partial \overline {{Q_1}} }}{{\partial p}} + {\rm{residue}}{{{\text{}} Q}}, $$ (6) where residueh, residuew, and residueQ are the residual terms. They express nonlinear interactions between subseasonal and nonsubseasonal timescales. The terms of residuew and residueQ contribute about 10% to their individual
$ {\partial {\rm PV}'}/{\partial t}$ over East Asia and its coastal region. The term of residueh is slightly higher than 10%. This study focuses on the subseasonal timescale. Thus, the three residual terms are ignored. Main contributors to Eqs. (4)–(6) are: the horizontal advection of subseasonal PV anomaly [${\rm{  }}(\bar u\;\partial {\rm{PV}}'\!/\partial x + \bar v\;\partial {\rm{PV}}'\!/\partial y)$ ] in Eq. (4), the vertical advection via$\omega '$ and the vertical distribution of climatological PV [${\rm{  }}\omega '(\partial \overline {{\rm{PV}}} /\partial p)$ ] in Eq. (5), and the vertical distribution of the subseasonal${Q_1}$ anomaly, i.e., [${{  g}}(f + \bar \zeta)\partial {Q'_1}/\partial p$ ] in Eq. (6), respectively.The procedure to obtain the subseasonal anomalies of every variable is as follows (Ren et al., 2015). Firstly, we use a fiveday running mean to remove very high frequency fluctuations in the daily data. Then, we discard the daily climatology; and lastly, the interannual timescale is eliminated via deducting seasonal anomaly. A multivariate empirical orthogonal function (MVEOF) analysis is used to obtain the spatiotemporal features of zonal oscillation of the WPSH on the subseasonal timescale. The MVEOF analysis is performed on normalized subseasonal anomalies in horizontal winds (
$u$ and$v$ ) at 500 hPa over the region (0°–45°N, 110°–150°E) for the period of early summer (1 June–20 July) from 1979 to 2016 (total 1900 days). In comparison with a regular EOF analysis, the MVEOF is more suitable to capture the spatial distributions of several selected variables and the covariability relationship among these variables. Timelagged regressions of the subseasonal variables on the standardized first principal components of MVEOF (PC1) are calculated. Day 0 means simultaneous regression. A negative lag day (a positive lag day) is performed by shifting backward (forward) the number of the leading (lag) days.