-
In this study, we used the seasonal to decadal climate prediction system IAP-DecPreS, which was developed by the State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG) in the Institute of Atmospheric Physics (IAP) of the Chinese Academy of Sciences (CAS) (Wu et al., 2015, 2018; Hu et al., 2019, 2020).
The IAP-DecPreS is composed of two parts, the coupled climate system models FGOLAS-f3-L, and the weakly coupled data initialization scheme EnOI-IAU (Wu et al., 2018). The FGOLAS-f3-L is one of three versions of the CAS model that have participated in the Coupled Model Intercomparison Project Phase 6 (CMIP6) (Zhou et al., 2020). The atmospheric component of FGOALS-f3-L is FAMIL (Zhou et al., 2015; Li et al., 2019), which is established on a finite volume dynamical core on a cubed-sphere grid with the horizontal resolution approximately equal to 1°×1°. The FAMIL has 32 layers in the vertical direction, with the top layer at 2.16 hPa (Guo et al., 2020a; He et al., 2020b). The oceanic component of FGOALS-f3-L is the low-resolution LICOM3 (Yu et al., 2018; Lin et al., 2020), which has 360 grid cells in the zonal direction and 218 grid cells in the meridional direction approximately equal to 1°×1°. To better resolve the equatorial waves, the meridional resolution refines from 1° to 0.5° near the equator. The low-resolution LICOM3 has 30 layers in the vertical direction, which is 10 m per layer in the upper 150 m and divided in uneven vertical layers below 150 m.
The FGOLAS-f3-L can reasonably reproduce the upper-level South Asian High, the low-level monsoonal circulations, the WPSH, and the mid-Pacific trough, and thus simulate the observed climatological vertical shear of meridional winds in July over the tropical western Pacific (Fig. 1). As mentioned in previous studies, the vertical shear of the climatological-mean meridional winds is vital for the baroclinic energy conversion from the basic flow to the perturbations associated with the PJ pattern (Kosaka and Nakamura, 2006). More details about the FGOALS-f3-L can be found in He et al. (2020a); Guo et al. (2020b); Guo et al. (2020a); He et al. (2020b). The EnOI-IAU assimilates the gridded SST from the HadISST version 1.1 (Rayner et al., 2003) and the subsurface temperature and salinity profiles from the EN4 dataset produced by the Hadley Centre (Good et al., 2013), which form the initial conditions for the FGOLAS-f3-L to conduct climate predictions.
Figure 1. Left panel: Observed climatological July mean (a) horizontal winds at the 150-hPa level (vectors, units: m s–1) and (c) precipitation (shaded, units: mm d–1) and horizontal winds at the 850-hPa level (vectors, units: m s–1) for the period 1995–2014. The observation derives from ERA5. Right panel: (b, d) as in (a, c), but for the FGOALS-f3-L historical simulations.
To investigate the predictability of the large-scale circulation anomalies over East Asia in July 2021, 21-member 3-month ensemble seasonal forecast runs were conducted from the end of June in 2021, with initial conditions derived from the outputs of the three initialization runs based on the EnOI-IAU. The time lagged method was utilized to generate the 21 prediction members. Here, seven start dates from 24 to 30 June were selected. We also used 3-member seasonal hindcast experiments for the period 2000–20 to remove the lead-time dependent model drifts, following the procedures recommended by Boer et al. (2016). The 3-month hindcast runs initialized from the end of June were conducted once per year for the period of 2000–20. More details of the experiment designs can be found in Table 1.
Exp name Integration Ensemble size Initial condition Initialization Jan 1950 to Dec 2021 3 Model states in 1 Jan 1950 from three historical runs Hindcast Initiated from the end of each month in each year during 2000–20, integrated 16 months 3 Model states in 25th, 28th, 30th of each month in each year derived from 3 initialization runs Forecast Initiated from the end of June in 2021, integrated 3 months 21 Model states in 24–30 June 2021 derived from 3 initialization runs Table 1. Designs of seasonal climate prediction experiments.
-
In this study, we used monthly and hourly data from the European Centre for Medium-Range Weather Forecasts reanalysis ERA5 with a horizontal resolution of 0.25°×0.25° (Hersbach et al., 2020). We also used the monthly sea surface temperature (SST) data derived from the HadISST version 1.1 at a horizontal resolution of 1°×1° (Rayner et al., 2003). All the datasets cover the period 1979–2021.
-
To understand the processes responsible for the precipitation anomalies over the tropical western Pacific, we diagnosed the atmospheric moisture equation (Seager et al., 2010; Seager et al., 2012; Chou et al., 2013) as follows:
where
$ P $ is precipitation,$ q $ is specific humidity,$ \boldsymbol{V} $ is horizontal wind,$ \omega $ is vertical pressure velocity, and$ p $ represents the vertical direction. The angle bracket$ < \mathrm{ }\mathrm{ }\mathrm{ }\mathrm{ } > $ denotes the mass-weighted vertical integral through the entire atmospheric column. The overbars (primes) represent the climatological monthly mean (monthly anomaly). The horizontal and vertical moisture advection terms can be divided into the thermodynamic components ($- < \bar{\boldsymbol{V}}\cdot {\nabla }_{h}q{'} > - < \bar{\omega }\cdot {\partial }_{p}q{'} >$ ), the dynamic components ($- < \boldsymbol{V}'\cdot {\nabla }_{h}\bar{q} > - < \omega'\cdot {\partial }_{p}\bar{q} >$ ), and the nonlinear components ($ \mathrm{N}\mathrm{L} $ ). The “Residual” denotes the residual term.To estimate the relative contributions of high- and low-frequency eddies to the precipitation anomalies at the monthly time scale, we also used the daily data to diagnose the column-integrated atmospheric moisture equation at intraseasonal (10−30 days) and synoptic (<10 days) timescales. Equation (1) can be rewritten as:
where the tilde represents the monthly mean. The superscript
$ * $ represents the intraseasonal (10–30 days) time scale, and the superscript '' represents the synoptic (<10 days) time scale. The intraseasonal and synoptic variations of each variable are obtained by the Lanczos filter. On the right hand side of the equation, the first (second) row is the relative contribution from the intraseasonal (synoptic) variability to the monthly precipitation anomalies. The third and fourth rows are contributions from the interactions between intraseasonal and synoptic variations and synoptic transient processes to the monthly precipitation anomalies, respectively, the sum of which are treated as the nonlinear components.The atmospheric moist static energy (MSE) equation (Neelin and Held, 1987; Wu et al., 2017) was diagnosed to investigate the processes driving the anomalous vertical motions, which is written as follows:
where
${C}_{p}$ and$ {L}_{v} $ are the specific heat at constant pressure and the latent heat of vaporization.$ T $ denotes the air temperature.$ q $ is the specific humidity.$ \varphi $ denotes the geopotential.$ u $ ,$ v $ , and$ \omega $ represent the zonal wind, meridional wind, and vertical pressure velocity, respectively, and$ x $ ,$ y $ , and$ p $ represent the zonal, meridional, and vertical direction, respectively.$ {C}_{p}T+{L}_{v}q $ is the atmospheric moist enthalpy, and$ h $ denotes the MSE, which is equal to${C}_{p}T+{L}_{v}q+\varphi$ .$ {F}_{\mathrm{n}\mathrm{e}\mathrm{t}} $ represents the net flux into the atmospheric column. The primes represent the monthly anomaly. The angle brackets denote the mass-weighted vertical integral through the entire atmospheric column. The MSE equation can be simplified towhere
$ \mathrm{N}\mathrm{L} $ is the nonlinear term. The bars (primes) represent the climatological monthly mean (monthly anomaly).The anomalous net flux into the atmospheric column (
${F}'_{\mathrm{n}\mathrm{e}\mathrm{t}}$ ) can be written as:where
$ {S}' $ represents the net shortwave radiative flux anomalies into the atmospheric column from the surface and the top of atmosphere. The net longwave radiative flux anomalies into the atmospheric column from the surface and the top of atmosphere can be separated into clear-sky and cloud-related components (${R}_{\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{r}}'$ and${R}_{\mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}}'$ ).$\mathrm{L}\mathrm{H}'$ represents the latent heat anomalies, and$\mathrm{S}\mathrm{H}'$ represents the sensible heat anomalies.The 3D wave-activity flux calculated in this study parallels that of Takaya and Nakamura (2001). The PJ pattern in July is defined as the leading singular vector decomposition (SVD) mode for the 500-hPa geopotential height field over East Asia (10°–70°N, 100°–160°E) and precipitation over the Philippines (10°–30°N, 125°–155°E), following Tao et al. (2017). Before the SVD analysis, both the precipitation anomalies and the 500-hPa geopotential height (Z500) anomalies are standardized and weighted by the square root of cosine of latitudes. SVD analysis also has been used in previous studies to define the PJ pattern at daily time scales (Sun et al., 2021).
Following Boer et al. (2018), the predictable variabilities in the forecast are estimated as the ensemble mean, while the unpredictable or “noise” components are the deviations of each member to the ensemble mean (ensemble spread), which are assumed to be independent among the ensemble members and to average to zero for a sufficiently large ensemble. In this study, we conducted the empirical orthogonal function (EOF) analysis to the ensemble spread to derive the leading modes of the unpredictable variabilities. To measure the intensity of the northward shift of the WPSH, we defined the Z850 dipole index, which is calculated based on the difference of geopotential height anomaly averaged over the low latitudes (20°–35°N, 110°–150°E) and midlatitudes (40°–55°N, 120°–160°E) of the East Asian coastal region. The precipitation anomalies averaged over the western tropical Pacific (15°–30°N, 125°–155°E) is defined as the WP–PR index. The Niño-3.4 index is the area-averaged SSTAs in 5°S–5°N, 170°–120°W. The student’s t test was used to estimate the significance of regression and correlation analysis.
-
To investigate the predictability of the northward shift of the WPSH in July 2021, we conducted 21-member 3-month ensemble seasonal forecasts initialized from the end of June in 2021 based on the IAP-DecPreS, which assimilates the observed oceanic data (including the SST and subsurface temperature and salinity profiles). The most important source of seasonal predictability stems from the oceanic variation, and thus, we first investigate the prediction of SSTA in July 2021 (Fig. 6). The forecasts from individual ensemble members are in agreement with each other and can predict the observed spatial distribution of SSTA well, with the uncentered pattern correlation coefficient (PCC) between the forecast and the observation ranging from 0.32 to 0.62. Compared with forecasts from individual members, the ensemble mean forecast has the highest prediction skill of PCC, and it reasonably reproduces the observed negative SSTA in the tropical central to eastern Pacific and positive SSTA in the eastern tropical Indian Ocean, the North Pacific, and the tropical South Atlantic (Fig. 6).
Figure 6. Spatial distributions of the observed and predicted global SSTA (units: °C) in July 2021. The observation and ensemble mean forecast are shown on the left and the right of the first row. Forecasts from individual members are shown below the first row. PCC is the uncentered pattern correlation coefficient. The forecasts by the IAP-DecPreS are started from the end of June in 2021.
In contrast, predictions of the 850-hPa atmospheric circulation over the Northern Hemisphere show spread among the ensemble members (Fig. 7). The prediction skill of individual members for this feature is lower than the prediction skill for the SSTA, with PCCs in 18 out of 21 members below 0.3 (Table 2). The ensemble mean, which is usually regarded as the predictable component in the forecasts, exhibits positive 850-hPa geopotential height anomalies over the midlatitude Far East, suggesting that the northward shift of the WPSH in July 2021 stems at least partly from the predictable component induced by the initial oceanic conditions. There are two realizations (#21 and #16) matching the observations closely, with PCC reaching 0.70 and 0.54, respectively, indicating that the unpredictable or “noise” component also plays a vital role. Hence, we focus on the respective impacts of the predictable and the unpredictable components of the northward shift of the WPSH in the following sections.
Figure 7. As in Fig. 5, but for the 850-hPa geopotential height anomalies (units: gpm) over the Northern Hemisphere.
$ {\mathrm{P}\mathrm{C}\mathrm{C}}_{\mathrm{S}\mathrm{S}\mathrm{T}} $ $ {\mathrm{P}\mathrm{C}\mathrm{C}}_{\mathrm{Z}850} $ $ {\mathrm{P}\mathrm{C}\mathrm{C}}_{\mathrm{S}\mathrm{S}\mathrm{T}} $ $ {\mathrm{P}\mathrm{C}\mathrm{C}}_{\mathrm{Z}850} $ Ens. Mean 0.62 0.29 Mem #11 0.53 0.26 Mem #1 0.32 −0.40 Mem #12 0.57 0.31 Mem #2 0.53 0.02 Mem #13 0.57 0.22 Mem #3 0.39 −0.07 Mem #14 0.55 0 Mem #4 0.45 −0.28 Mem #15 0.48 0.25 Mem #5 0.43 0.21 Mem #16 0.57 0.54 Mem #6 0.52 −0.21 Mem #17 0.56 0.25 Mem #7 0.35 −0.25 Mem #18 0.51 0.08 Mem #8 0.48 0.08 Mem #19 0.42 0.19 Mem #9 0.49 0.07 Mem #20 0.41 −0.20 Mem #10 0.56 0.13 Mem #21 0.62 0.70 Table 2. The uncentered pattern correlation coefficient (PCC) between the forecasts and the observations for 850-hPa geopotential height field and SSTA in Figs. 6 and 7.
-
Figure 8 shows the predictable variability of the large-scale circulation and precipitation anomalies in July 2021, which are derived from the 21-member ensemble mean of the forecasts by IAP-DecPreS. There are positive geopotential height anomalies over northeastern East Asia at both the 500-hPa and 850-hPa levels (Figs. 8a, b), suggesting that part of the variability of the WPSH in July 2021 is predictable, and the IAP-DecPreS can predict the northward shift of the WPSH at a 1-month lead. The predictability stems from the positive precipitation anomalies over the South China Sea and the tropical western Pacific, which excite a PJ pattern with cyclonic (anticyclonic) anomalies over the South China Sea and the tropical western Pacific (northeastern East Asia), respectively (Fig. 8c). The relative vorticity anomalies with a poleward phase tilt with height and the northward propagation of the Rossby waves are consistent with the observations. The prediction shows a northward propagation of Rossby waves in the upper troposphere, instead of the observed southeastward propagation (Figs. 5a and 8d). The predictable PJ pattern is confined over the coastal regions of East Asia, and the intensity of the northward shift of the WPSH is weaker than that seen in the observations. The predicted Z850 dipole index in the ensemble mean prediction is 10.3 gpm, which is about 28.0% of the observed value.
Figure 8. The 21-member ensemble mean predicted (a) 500-hPa geopotential height anomalies (units: gpm), (b) 850-hPa geopotential height anomalies (units: gpm), (c) precipitation (shaded, units: mm d−1) and 850-hPa horizontal wind (vector, units: m s−1) anomalies, and (d) the meridional sections of relative vorticity (shaded, units: 10−6 s−1) and TN-flux (vector, units: m2 s−2, the vertical components have been multiplied by 100) averaged between 110°–130°E in July 2021.
What is the mechanism of the predictable signal? We diagnosed the moisture equation [Eq. (1)] to the precipitation anomalies over the tropical western Pacific (10°–25°N, 125°–150°E). The results show that the positive precipitation anomalies are caused by the positive anomalous advection of the climatological vertical moisture by ascending anomalies (
$- < {\omega }'\cdot {\partial }_{p}\bar{q} >$ ) (Fig. 9a). In the tropical regions with deep convection, the gross moist stability is usually positive and the$ {\partial }_{p}\stackrel{-}{h} $ is usually less than 0 (Back and Bretherton, 2006), and thus, the positive terms on the right-hand side of Eq. (4) can drive anomalous ascending motion to keep the MSE budget balance (Biasutti et al., 2018). MSE budget analysis over the tropical western Pacific (10°–25°N, 125°–150°E) indicated that the$< {\omega }'{\partial }_{p}\bar{h} >$ term is primarily balanced by positive net energy flux anomalies (${F}_{\mathrm{n}\mathrm{e}\mathrm{t}}'$ ) and the horizontal advection of climatological enthalpy by wind anomalies ($- < {u}'{\partial }_{x}\overline{\left({C}_{p}T+{L}_{v}q\right)} >$ and$- < {v}'{\partial }_{y}\overline{\left({C}_{p}T+{L}_{v}q\right)} >$ ), which represent two different physical processes.Figure 9. Budget analysis for (a) the moisture equation [Eq. (1), units: mm d–1] and (b) the MSE equation [Eqs. (4–5), units: W m–2] for the area 10°–25°N, 125°–150°E in July 2021 from the 21-member ensemble mean forecasts by IAP-DecPreS.
The
${F}_{\mathrm{n}\mathrm{e}\mathrm{t}}'$ is determined by the cloud-related longwave radiative flux anomalies (${R}_{\mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}}'$ ) and the surface latent heat anomalies ($\mathrm{L}\mathrm{H}^{'}$ ). The${R}_{\mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}}'$ is generated by internal positive feedback between convection and cloud radiative forcing in the tropical atmosphere (Bretherton and Sobel, 2002; Su and Neelin, 2002; Neelin and Su, 2005). When the convection over the tropical western Pacific is enhanced, the deep convective clouds and associated cirrostratus and cirrocumulus will increase, which leads to a net warming of the atmospheric column, a further weakening of the gross moist stability of the atmospheric column, and thus an enhancement of the ascending flow and convective activities (Fig. 10a). The positive$ \mathrm{L}\mathrm{H}^{'} $ anomalies are associated with the strengthened low-level summer monsoon circulations, which enhance surface evaporation and thus heat the atmospheric column and lead to ascending anomalies (Fig. 10b). The terms$- < {u}'{\partial }_{x}\overline{\left({C}_{p}T+{L}_{v}q\right)} >$ and$- < {v}'{\partial }_{y}\overline{\left({C}_{p}T+{L}_{v}q\right)} >$ are dominated by the anomalous horizontal advections of climatological latent heat energy by wind anomalies (Figs. 10c, d), which are more prominent at the low levels (not shown). At the 925-hPa level, the July climatological specific humidity over the tropical western Pacific is greater than the surroundings due to the outbreak of the western Pacific summer monsoon, which induces the positive meridional gradient of climatological specific humidity over the subtropical regions (Fig. 10e). The negative convective heating associated with the La Niña induces a pair of anomalous anticyclones over the tropical Pacific, one in both the Northern and Southern Hemisphere. There are southwesterlies at the west edge of the northern anticyclone, which strengthen the monsoonal circulations and induce the positive moisture enthalpy advection anomalies over the tropical western Pacific (Fig. 10e).Figure 10. Spatial distributions of (a) the cloud-related longwave radiative flux anomalies (
${R}_{\mathrm{c}\mathrm{l}\mathrm{o}\mathrm{u}\mathrm{d}}'$ ), (b) the surface latent heat anomalies ($\mathrm{L}\mathrm{H}^{'}$ ), (c) the zonal advection of climatological enthalpy by wind anomalies ($- < {u}'{\partial }_{x}\overline{\left({C}_{p}T+{L}_{v}q\right)} >$ ), and (d) the meridianal advection of climatological enthalpy by wind anomalies ($- < {v}'{\partial }_{y}\overline{\left({C}_{p}T+{L}_{v}q\right)} >$ ). (e) the July 925-hPa climatological specific humidity (shading; units: kg kg–1) and 925-hPa horizontal wind anomalies in the 21-member ensemble mean predictions by IAP-DecPreS.The above analysis demonstrates that the predictable component of the northward shift of the WPSH is related to the PJ pattern, which originates from the tropical circulation anomalies associated with the La Niña event and associated local positive cloud–radiation and wind–SST–evaporation feedbacks.
-
The predictable component only contributes 28.0% of the observed WPSH variation in July 2021, and thus, the remaining 72.0% of the variation is associated with the unpredictable component. The leading modes of the unpredictable variability of the circulation anomalies are obtained by performing EOF analysis on the ensemble spread of the 500-hPa geopotential height anomalies over northeastern East Asia (20°–60°N, 110°–170°E). The first two EOF modes account for 49.5% and 17.4% of the total variance, respectively. The first EOF (EOF1) shows a meridional tripole pattern of geopotential height anomalies over the coastal regions of East Asia, with negative anomalies over the northern part of the Philippine Sea and the Bering Strait and positive anomalies over the northern part of Japan (Fig. 11a). The second EOF (EOF2) exhibits more prominent geopotential anomalies over the midlatitude to high-latitude regions, with weak positive anomalies to the east of Japan (Fig. 11b). Relationships between the normalized principal components corresponding to the first two EOF modes (PC1 and PC2) and the ensemble spread of the Z850 dipole index are shown in Figs. 11b and d. The predicted Z850 dipole index in the ensemble members ranges from –53.4 gpm to 62.2 gpm, which covers the observed value of 36.8 pgm. EOF1 has a closer correlation relationship with the original prediction than EOF2, with the correlation coefficient reaching 0.99 (significant at the 99% level). The agreement between the observed Z850 dipole index and the predictions in the members with greater loading of EOF1 indicates that the combination of the predictable component and the extreme positive phase of EOF1 can reproduce the observed northward shift of the WPSH in July 2021. The low correlation coefficient between PC2 and the ensemble spread of the Z850 dipole index indicates that EOF2 is not the main factor dominating the unpredictable variability (Fig. 11d), and we further verify that EOF2 is associated with the midlatitude wave train over Eurasia (not shown). In the following analysis, we focus on the formation mechanism of EOF1.
Figure 11. Spatial distributions of the 850-hPa geopotential height anomalies (units: gpm) associated with (a) the first EOF mode (EOF1) and (c) the second EOF mode (EOF2) of the ensemble spread of the forecasts. Variance contributions of the two EOF modes are noted in the parentheses. Relationships between predicted Z850 dipole index and (b) the normalized principal component (PC) corresponding to the first EOF mode (PC1) and (d) the normalized PC corresponding to the second EOF mode (PC2). The correlation coefficients between PC1 (PC2) and the predicted Z850 dipole index are shown in the legends. The black line in (b) is the linear fitting equation between PC1 and the predicted Z850 dipole index. The blue triangle represents the observation, for which the PC value is estimated through the linear fitting equation in the model world. The hatching in (a, c) denotes values exceeding the 95% confidence level.
The circulation, precipitation, and SST anomalies associated with EOF1 are shown in Fig. 12. EOF1 exhibits a typical PJ pattern, with the upper-level circulation anomalies shifted meridionally relative to the low-level anomalies by about a quarter wavelength and northward (southeastward) propagation of the Rossby waves in the lower (upper) troposphere (Figs. 12a, c, e). There are no significant SSTAs associated with EOF1 except for those in the northern Pacific, which are below the centers of the PJ pattern-related anomalous atmospheric activity (Fig. 12b), indicating that EOF1 is independent of the SSTA’s forcing. The PJ pattern of EOF1 is closely related to the monthly precipitation anomalies over the tropical western Pacific (Fig. 12d), which was suggested to be internally generated by both the intraseasonal and synoptic variabilities (Kawamura and Ogasawara, 2006; Yamada and Kawamura, 2007; Wang et al., 2016).
Figure 12. Regressions of the ensemble spread of (a) relative vorticity (shaded; units: m2 s–2) and TN-flux (vector, units: m2 s–2) anomalies at 200 hPa, (b) SSTA (units: °C), (c) relative vorticity (shaded; units: m2 s–2) and TN-flux (vector, units: m2 s–2) anomalies at 850 hPa, (d) precipitation anomalies (units: mm d–1), (e) the meridional sections of relative vorticity (shaded, units: 10–6 s–1) and TN-flux (vector, units: m2 s–2, the vertical components have been multiplied by 100) averaged between 110°–130°E, and (f) 850-hPa eddy kinetic energy (units: m2 s–2) at a time scale of 10–30 days onto the normalized principal component (PC) corresponding to the first EOF mode (PC1) of the ensemble spread of the forecasts in July 2021. The hatching denotes values exceeding the 95% confidence level. The triangles in (a, c, d, f) represent the center of precipitation anomaly over the western tropical Pacific.
To determine what kind of internal variability induces the EOF1-related precipitation anomalies over the tropical western Pacific, we diagnosed the column-integrated atmospheric moisture equation at intraseasonal (10–30 d) and synoptic (<10 d) time scales based on the daily data. The moisture budget associated with EOF1 indicates that the precipitation anomalies over the tropical western Pacific at the monthly time scale are dominated by the intraseasonal variability (Fig. 13). This is further confirmed by the significant 850-hPa eddy kinetic energy at a time scale of 10–30 days over the tropical western Pacific associated with EOF1 (Fig. 12f). Hence, different from the La Niña-forced PJ pattern, the PJ pattern related to EOF1 here is generated by the atmospheric internal processes due to intraseasonal variability.
Figure 13. Regressions of the ensemble spread of moisture budgets onto the normalized principal component (PC) corresponding to the first EOF mode (PC1) of the ensemble spread of the forecasts in July 2021 at time scales of (a–d) monthly, (e–h) 10–30 days, and (i–l) less than 10 days. (a, e, i) precipitation anomalies and (b, f, j) the dynamic components, (c, g, k) the thermodynamic components, and (d, h, l) the nonlinear components of the anomalous moisture advection. The hatching denotes values exceeding the 95% confidence level.
Exp name | Integration | Ensemble size | Initial condition |
Initialization | Jan 1950 to Dec 2021 | 3 | Model states in 1 Jan 1950 from three historical runs |
Hindcast | Initiated from the end of each month in each year during 2000–20, integrated 16 months | 3 | Model states in 25th, 28th, 30th of each month in each year derived from 3 initialization runs |
Forecast | Initiated from the end of June in 2021, integrated 3 months | 21 | Model states in 24–30 June 2021 derived from 3 initialization runs |