The interdecadal variations of cross-month correlation feature of the NWPSH

Abstract

The Northwest Pacific Subtropical High (NWPSH) presents a notable and even counterintuitive phenomenon: it shows a strong positive correlation between June and August (correlation coefficient: 0.55, significant at the 99.5% confidence level) during 1979–2005, while this correlation weakens sharply to -0.05 in 2006–2024. However, the relationships between June and July, and between July and August remain consistently weak throughout the entire period. Positive sea surface temperature (SST) anomalies in the tropical North Indian Ocean (TNIO) contribute to a persistent intensification of the NWPSH from June to August before 2005. Meanwhile, the Boreal Summer Intraseasonal Oscillation (BSISO) exhibits a stronger 60-day periodicity, which can cause opposite (similar) variations of NWPSH in adjacent month (cross-month). Under the combined effects of TNIO SST positive anomalies and BSISO, the NWPSH mainly exhibits the cross-month correlation feature, while the relationships between adjacent months are very weak. After 2005, the BSISO exhibits a marked shortening of its periodicity, and the key SST regions associated with the NWPSH in June, July, and August are also inconsistent, which induce the weakening of the cross-month correlation feature of the NWPSH. The shortening of the BSISO’s periodicity is attributed to warming over the Maritime Continent and the tropical western Indian Ocean, which intensify both the zonal Walker and meridional Hadley circulations. These changes enhance downward motions over the tropical eastern Indian Ocean and the northwestern Pacific, thereby suppressing the initiation of BSISO’ convection and accelerating its decay.

Introduction

The Northwest Pacific Subtropical High (NWPSH) is located in the northwest Pacific region, encompassing a vast subtropical area and adjacent to the east Asia continent. It is influenced by multiple factors, including solar radiation, land-sea thermal contrast, mid-to-high latitude circulations, Indo-Pacific sea surface temperature (SST), and boreal summer intraseasonal oscillation (BSISO). As a result, the NWPSH serves as a crucial circulation system linking the tropical and mid-latitude, as well as oceanic and continental climates^1,2,3,4,5. Owing to the impacts of these factors across distinct temporal scales, the NWPSH demonstrates pronounced variations spanning from intraseasonal to interdecadal timescales. Therefore, it has important impacts on the climate anomalies over east Asia across these time scales^{6,7,8,9,10,11}.

Although the NWPSH exhibits significant variabilities from intraseasonal to interdecadal time scales, current researches on the NWPSH mainly focus on a single time scale^{12,13,14,15,16,17}. Several studies have indicated that the summer-averaged NWPSH exhibits interdecadal variations, which are primarily reflected in two aspects: (1) The interdecadal variations of key SST regions associated with the summer NWPSH, as well as the interannual periodicity of the summer NWPSH itself, are closely linked to the interdecadal variations in the El Niño–Southern Oscillation (ENSO) after 2000^5,18,19. (2) Since the late 1970s, the summer NWPSH shows interdecadal variations characterized by enhanced intensity, expanded coverage, and a westward displacement, which are believed to be associated with global warming^20,21,22.

However, studies on the interdecadal variations of the intraseasonal variabilities of the NWPSH are relatively scarce. As a key intraseasonal circulation system in the mid-to-low latitudes, the BSISO has an active center located in the northwestern Pacific^23,24,25,26. Therefore, the BSISO can exert significant influences on the intraseasonal variabilities of the NWPSH^27,28,29. The BSISO also strongly influences the intraseasonal variabilities of the northwestern Pacific summer circulation, driving a transition from strong anticyclonic circulation in June to pronounced cyclonic circulation by August in 2016³⁰. In addition, the first four leading modes of the BSISO are characterized by significant positive or negative outgoing longwave radiation (OLR) anomalies over the northwest Pacific, which favor the strengthening or weakening of the NWPSH^9,31,32. Relevant studies have indicated that the summer La Niña and El Niño SST patterns can exert different influences on the vertical shear of zonal winds and moisture anomalies in the tropics, which will lead to changes in the intensity, periodicity, and propagation pathways of the BSISO^33,34,35,36. However, whether the interdecadal variations in ENSO around 2000 have led to interdecadal variations in the impacts of BSISO on the NWPSH remains unclear.

Therefore, this study investigates the interdecadal variations in the influence of the BSISO on the intraseasonal variabilities of the NWPSH, as well as the relationships between the NWPSH in different summer months, in the context of the interdecadal shifts of ENSO around 2000. It is crucial for gaining a deeper understanding of the interactions among various circulation systems in east Asia, uncovering the physical mechanisms of intraseasonal drought and flood disasters, and providing a scientific basis for disaster prevention and mitigation efforts.

Results

Interdecadal Variations of the Cross-month Correlation Feature of the NWPSH and Their Impacts

Figure 1a–d show the spatial distribution of the standard deviation of the 850hPa geopotential height field for June, July, August, and the summer mean, respectively. Overall, the locations of the centers of standard deviation are generally consistent across different months and the summer mean. These centers are mainly located near and south of 25°N, exhibiting a southwest-northeast orientation. However, the standard deviation for the individual months is significantly larger than that of the summer mean, and the standard deviation of the geopotential height field gradually increases from June to August.

Fig. 1: The relationship of the NWPSH across different months.

a–d The standard deviation of the 850hPa geopotential height field (unit: ${\rm{gpm}}$) for June, July, August, and the summer mean, respectively. The purple box denotes the region for defining the NWPSH index. e The 21-year sliding correlation of the NWPSH index between June and August (purple solid line), June and July (red solid line), and July and August (blue solid line). Horizontal (vertical) gray dashed line indicates the 95% confidence level threshold for the 21-year sliding correlation (2005, the dividing year for the two periods). The purple numbers in the upper left (upper right) corner show the correlation coefficient and confidence level between the June and August NWPSH indices during 1979–2005 (2006–2024).

A 21-year sliding correlation between the NWPSH indices in different summer months reveals a notable phenomenon. Before 2005, there is a strong correlation between the NWPSH in June and August, which drops significantly thereafter. The correlation coefficient between June and August NWPSH during 1979–2005 is 0.55 (significant at the 99.5% confidence level), while it drops to –0.05 during 2006–2024. The correlation between the June and July NWPSH transitions from a weak positive to a weak negative relationship, and then reverts to a weak positive correlation after 2000. Although the correlation between the July and August NWPSH gradually increases after 2000, this trend is not statistically significant and slightly weakens after 2005 (Fig. 1e). Therefore, the NWPSH exhibits a distinct cross-month correlation feature, in which the NWPSH in June, skipping July, is significantly correlated with that in August. But this feature notably weakens after 2005. Both Fisher r-to-z transformation and permutation tests methods indicate that the difference in correlation coefficients (0.55 for 1979–2005 vs. –0.05 for 2006–2024) is statistically significant at the 95% confidence level.

Before 2005, the NWPSH index in June is not only closely related to the 850hPa geopotential height field in June, with significant positive anomalies extending southeastward from central Asia to the tropical central Pacific, and a high-value center located in the northwestern Pacific (Fig. 2a). The June NWPSH index is also closely associated with the 850hPa geopotential height field in August, with positive anomalies extending from central Asia southeastward to the northwestern Pacific (Fig. 2c). However, the relationship between the June NWPSH index and the July 850hPa geopotential height field is relatively weak. No prominent anomalies are observed over the Asian continent or the northwestern Pacific, except for some significant positive anomalies over the North Pacific (Fig. 2b). This further confirms that the NWPSH shows weak connections between adjacent months but exhibits a notable cross-month correlation feature before 2005. In contrast, the linkage between the June NWPSH index and the July 850hPa geopotential height field becomes stronger after 2005 (Fig. 2e). However, the relationship between the June NWPSH index and the August 850hPa geopotential height field becomes very weak (Fig. 2f). Similarly, the regressions of the August NWPSH index onto the 850hPa geopotential height fields in June, July, and August yield consistent results (Figure omitted).

Fig. 2: The interdecadal variations in the impact of the June NWPSH on the circulation.

a–c The regressions of the June NWPSH index onto the 850hPa geopotential height field (unit: ${\rm{gpm}}$) for June, July, and August in 1979–2005. The purple box denotes the region for defining the NWPSH index. d–f are similar to (a–c), but for 2006-2024. Dots indicate significance at the 95% confidence level.

The NWPSH is one of the most important atmospheric circulation systems over the northwestern Pacific and east Asia, exerting significant influences on summer precipitation^9,10,19. Since there is the significant cross-month correlation feature of the NWPSH between June and August before 2005, it is likely that east Asian circulation and precipitation also exhibit similar connections. The June NWPSH not only induces pronounced anticyclonic water vapor transport and divergence over the northwestern Pacific, accompanied by strong moisture convergence over eastern China and the Maritime Continent (Fig. 3a), but also corresponds to similar patterns in August (Fig. 3c). In contrast, the June NWPSH does not show clear association with water vapor transport anomaly in July over the northwestern Pacific and east Asia (Fig. 3b). Therefore, the June NWPSH not only significantly reduces (increases) precipitation over the northwestern Pacific (eastern China and the Maritime Continent) in June, but also exerts similar influences in August (Fig. 4a; c). However, its impacts on July precipitation are minimal (Fig. 4b). After 2005, the relationship between the June and August NWPSH weakens considerably (Fig. 1e; Fig. 2f), resulting in marked decline in the influences of the June NWPSH on the water vapor transport and precipitation in August (Fig. 3f; Fig. 4f).

Fig. 3: The interdecadal variations in the impact of the June NWPSH on the circulation.

a–c The regressions of the June NWPSH index onto the vertically integrated water vapor flux (vectors; units: 10² kg m^-1 s^-1) and water vapor convergence/divergence anomalies (shading; units: 10⁵ gk m^-2 s^-1) from 1000 to 300 hPa for June, July, and August in 1979–2005. d–f are similar to (a–c), but for 2006–2024. Dots indicate regions significant at the 95% confidence level. Only vectors significant at the 95% confidence level are shown.

Fig. 4: The interdecadal variations in the impact of the June NWPSH on the precipitation.

a–c The regressions of the June NWPSH index onto the precipitation (units: mm day^-1) for June, July, and August in 1979–2005. d–f are similar to (a–c), but for 2006–2024. Dots indicate regions significant at the 95% confidence level.

The Physical Mechanisms of the Interdecadal Variations in the Cross-month Correlation Feature of the NWPSH

Why does the NWPSH exhibit intriguing cross-month correlation feature, along with evident interdecadal variations in this feature? We first investigate the variations in SST associated with the NWPSH in different months. Before 2005, the NWPSH in June, July, and August is significantly correlated with positive SST anomalies in the TNIO. There are no significant SST anomalies across the Pacific, except for weak positive anomalies in the tropical Pacific east of 135°W (Fig. 5a–c). After 2005, the tropical SST shows no clear association with the June NWPSH (Fig. 5d). The July NWPSH is mainly associated with positive SST anomalies in the Maritime Continent and negative SST anomalies in the central–eastern equatorial Pacific, while the August NWPSH is primarily influenced by negative SST anomalies in the central–eastern tropical Pacific (Fig. 5e–f). The positive SST anomalies in the TNIO excite Kelvin waves and equatorial easterly anomalies by humidifying and heating the lower atmosphere. These processes induce anticyclonic wind shear and downward motions over the northwestern Pacific, thereby strengthening the NWPSH^1,37. Before 2005, positive SST anomalies in the TNIO from June to August favor the persistent intensification of the NWPSH (Fig. 6), thereby leading to a strong connection between the NWPSH in June and August (Fig. 1e). After 2005, the key SST regions influencing the NWPSH in different months are inconsistent, resulting in weak relationships between the NWPSH in different months. During 1979–2005, not only do the positive SST anomalies in the TNIO continuously enhance the NWPSH from June to August, but the SST anomalies in the TNIO across different months are also closely linked (Table 1). Why are the NWPSH in June and August closely correlated, while the relationships of the NWPSH between June and July and between July and August are weak (Fig. 1e)?

Fig. 5: The interdecadal variations in SST associated with the NWPSH.

a–c The SST anomalies (shading, unit: ${\rm{K}}$) regressed onto the NWPSH index in June, July and August during 1979–2005. d–f are similar as (a–c), but for 2006–2024. Dots indicate significance at the 95% confidence level, and the red box denotes the region for defining the TNIO SST index.

Fig. 6: The interdecadal variations in the impact of TNIO SST on the NWPSH.

a–c 850hPa geopotential height field (shading, unit: ${\rm{gpm}}$) and horizontal wind field (vectors, unit: ${\rm{m}}{\cdot{\rm{s}}}^{-1}$) regressed onto the TNIO SST index in June, July and August during 1979–2005. Dots indicate significance at the 95% confidence level, and only vectors reaching the 95% confidence level are displayed.

Table 1 Correlation coefficients between TNIO SST index in different summer months, and between TNIO SST and NWPSH index during 1979-2005

Since SST is a low-frequency and slowly varying variable, its changes on intraseasonal timescales are relatively small (Fig. 5), making it insufficient to explain the cross-month correlation feature between the NWPSH in June and August. Hence, further investigation into the influences of higher-frequency factor, such as the BSISO, is essential to elucidate the distinct intraseasonal behavior of the NWPSH. Both EOF1 and EOF2 of the BSISO exhibit significant anticyclone and positive OLR anomalies over the northwestern Pacific, which are favorable for the enhancement of the NWPSH (Fig. 7a, b)^9,31. The relationships between BSISO’s PC1 and PC2 across different months remain unclear, and whether these relationships can help explain the cross-month correlation feature of the NWPSH also requires further investigation.

Fig. 7: The relationship of the BSISO across different months.

a and b daily OLR (shading, unit: W·m^-2) and 850 hPa horizontal wind field (vectors, unit: m·s^-1) regressed onto PC1 and PC2 of BSISO. The purple box denotes the region for defining the NWPSH index. Dots indicate significance at the 95% confidence level, and only vectors reaching the 95% confidence level are displayed. c 21-year sliding correlation of the June and August monthly averaged $\overline{{\rm{PC}}1}$ (purple solid line), June and July monthly averaged $\overline{{\rm{PC}}1}$ (red solid line), and July and August monthly averaged $\overline{{\rm{PC}}1}$ (blue solid line) of BSISO. Horizontal (vertical) gray dashed lines indicate the 95% confidence level threshold for the 21-year sliding correlation (2005, the dividing year for the two periods). Purple numbers in the upper left (upper right) corner show the correlation coefficient and confidence level of the June and August monthly averaged $\overline{{\rm{PC}}1}$ of BSISO during 1979-2005 (2006-2021). d is similar as (c), but for the monthly averaged $\overline{{\rm{PC}}2}$ of BSISO.

First, daily BSISO’s PC1 and PC2 values are averaged for June, July, and August to obtain the corresponding monthly means, denoted as $\overline{{\rm{PC}}1}$ and $\overline{{\rm{PC}}2}$. Then, 21-year sliding correlation is calculated between any two months of$\,\overline{{\rm{PC}}1}$ and $\overline{{\rm{PC}}2}$. The results show that the correlation between June and August $\overline{{\rm{PC}}1}$ exhibits a significant interdecadal weakening during 1979–2021. Specifically, the correlation coefficient is 0.52 during 1979–2005 (significant at the 99% confidence level), but drops to –0.29 during 2006–2021 (Fig. 7c). In contrast, the correlations between June and July $\overline{{\rm{PC}}1}$, as well as between July and August, are significantly negative throughout the entire period. For $\overline{{\rm{PC}}2}$, the correlation between June and August is also relatively stronger during 1979–2005 (0.46, significant at the 98% confidence level), but becomes very weak in 2006–2021 (Fig. 7d). In addition, the correlations of $\overline{{\rm{PC}}2}$ between June and July, as well as between July and August, are also significantly negative throughout the entire period. Therefore, during 1979–2005, under the influences of the BSISO, the NWPSH tends to exhibit opposite changes between June and July, and between July and August, but similar changes between June and August.

From June to August, the influences of the TNIO SST anomalies on the NWPSH remain stable (Fig. 5; Fig. 6). When the low-frequency and slowly varying forcings of the SST anomalies are superimposed with the high-frequency influences of BSISO, which tend to weaken the relationships between adjacent-month NWPSH and enhance the positive correlation between the NWPSH in June and August. As a result, during 1979–2005, the NWPSH primarily exhibits a strong cross-month correlation between June and August, whereas the correlations between adjacent months are relatively weak. In contrast, during 2006–2021, the correlation between the NWPSH in June and August weakens. This can be attributed to the reduced correlations of the BSISO’s $\overline{{\rm{PC}}1}$ and $\overline{{\rm{PC}}2}$ between June and August (Fig. 7c–d), as well as the absence of consistent SST anomalies during these two months (Fig. 5d–f). Overall, the cross-month correlation feature of the NWPSH is jointly modulated by both BSISO and TNIO SST anomalies.

Why do the cross-month positive correlation features of the NWPSH and the BSISO’s $\overline{{\rm{PC}}1}$ and $\overline{{\rm{PC}}2}$ occur in June-August? The answers may lie in the interdecadal variations in the intraseasonal periodicity of the NWPSH and the BSISO. To further extract the intraseasonal variabilities of the NWPSH, the daily NWPSH index is subjected to a 20-100-day Lanczos band-pass filtering, which is denoted as L-NWPSH. Wavelet analysis is subsequently applied to the L-NWPSH to investigate interdecadal variations in its intraseasonal periodicity. Compared to 2006–2021, the NWPSH exhibits a stronger 60-day periodicity during 1979–2005 (Fig. 8a). The differences in wavelet power spectrum between the two periods further reveal that this 60-day periodicity is primarily active from June to August during 1979–2005 (Fig. 8b). During 1979–2005, BSISO’s PC1 also exhibits a stronger 60-day periodicity, primarily occurring in June, July, and August (Fig. 8c, d). No clear interdecadal variation in the periodicity of BSISO’s PC2 is observed during 1979–2005 (Fig. 8e, f). However, the mean wavelet power spectrum of BSISO’s PC2 also shows a pronounced 60-day periodicity from June to August during 1993–2005 (Fig. 8g, h).

Fig. 8: The interdecadal variations in the wavelet power spectrum of the NWPSH and BSISO.

a Wavelet power spectrum of L-NWPSH, the purple vertical dashed line denotes the year 2005. The red, green, and blue curves represent the averaged power of 1979−2005, 2006−2021, and 1979−2021, respectively. b Differences in wavelet power spectrum of L-NWPSH between 1979−2005 and 2006−2021 (1979−2005 minus 2006−2021). c and d are similar to (a) and (b), but for PC1 of BSISO. e and f are similar to (a) and (b), but for PC2 of BSISO. g is similar to (a), but for PC2 of BSISO, with purple vertical dashed lines indicating the years 1993 and 2005. h is similar to (b), but for the red curve denotes the averaged power for 1993−2005, the green curve denotes the averaged power of 1979−1992 and 2006−2021, and the blue curve denotes the averaged power of 1979−2021. h is similar to (b), but for the differences in wavelet power spectrum of BSISO PC2 between 1993−2005 and other years (1993−2005 minus the other years). Dots indicate significance at the 95% confidence level.

To further investigate the causes of the interdecadal variations in the intraseasonal periodicity of the BSISO and NWPSH, this study further composites the evolutions of OLR anomalies associated with the intraseasonal variabilities of the NWPSH. The OLR field is reconstructed based on BSISO’ EOF1 and EOF2, and then the 10-day mean evolutions of OLR anomalies from day -39 to day +40 are composited, with day 0 corresponding to the peak convection over the northwest Pacific. A total of 46 BSISO events are selected, including 32 events during 1979–2005 and 14 events during 2006–2021.

In the climatology, pronounced negative OLR anomalies over the tropical Indian Ocean and positive anomalies over the northwestern Pacific are already evident during day –39 to –30, with their intensities further amplifying during day –29 to –20 (Fig. 9a, b). Subsequently, the negative OLR anomalies over the tropical Indian Ocean propagate northward and eastward, accompanied by the development of significant negative OLR anomalies over the Maritime Continent (Fig. 9c). During day -9 to 0, strong negative OLR anomalies are observed over the northwest Pacific, while positive anomalies dominate the tropical Indian Ocean (Fig. 9d). Starting from day +1, the negative anomalies over the northwest Pacific and the positive anomalies over the tropical Indian Ocean begin to weaken. By day +11 to +20, pronounced negative and positive OLR anomalies reemerge over the tropical Indian Ocean and northwestern Pacific, respectively, indicating the end of the prior BSISO event and the onset of the next (Fig. 9e, f). The OLR anomalies continue to propagate in a similar manner during the following phases (Fig. 9g, h).

Fig. 9: The influence of the BSISO on the intraseasonal evolution of the NWPSH.

a–h Evolutions of OLR anomalies (units: ${{\rm{W}}{\cdot} {\rm{m}}}^{-2}$) reconstructed by the BSISO’s EOF1 and EOF2, averaged over 10-day intervals from day –39 to day 40 during 1979–2021. Only OLR anomalies reaching the 95% confidence level are displayed.

The evolution of the abnormal spatial distribution of OLR shows little difference between the two periods (Figs. S1; S2). However, the differences in the intensity between the two periods can explain why the NWPSH and BSISO exhibit a stronger 60-day periodicity during 1979–2005. Compared with 2006-2021, the tropical Indian Ocean and the northwestern Pacific show stronger OLR negative anomalies and positive anomalies from day -39 to -20 during 1979-2005, which indicate the convective initiation of BSISO is earlier and stronger before 2005 (Fig. 10a, b). During day -19 to 0, as the climatology OLR anomalies over the northwestern Pacific shift from positive to negative (Fig. 9c, d), the propagation process of the BSISO remain generally consistent across the two periods (Fig. 10c, d). However, significant differences reemerge during the decay phases of the BSISO. In particular, during 1979–2005, more pronounced negative OLR anomalies persist over the northwest Pacific from day 1 to 20 (Fig. 10e, f). However, in the climatological mean evolutions of the BSISO, OLR anomalies over the northwestern Pacific have already shifted from negative to positive after day 11 (Fig. 9f). Therefore, compared to 2006–2021, the BSISO exhibits stronger and more prolonged convection over the northwestern Pacific in the decay phases during 1979–2005. To sum up, the earlier onset and slower decay of BSISO convection during 1979–2005 eventually lead to a longer BSISO periodicity.

Fig. 10: The influence of the BSISO on the intraseasonal evolution of the NWPSH.

a–h Evolutions of the differences between 1979–2005 and 2006–2021 (1979–2005 minus 2006–2021) of the OLR anomalies (units: W m^-2). The anomalies are reconstructed by the BSISO’s EOF1 and EOF2, averaged over 10-day intervals from day -39 to day 40. Only OLR anomalies reaching the 95% confidence level are displayed.

As demonstrated above, the BSISO associated with intraseasonal variabilities of the NWPSH exhibits an earlier onset and slower decay during 1979–2005. These characteristics ultimately result in a longer periodicity of both the BSISO and NWPSH before 2005. The interdecadal variations in the periodicity of the BSISO can be attributed to the interdecadal enhancement of the Walker and Hadley circulations, which are induced by changes in the SST background state^38,39,40.

Compared with the period of 1979–2005, the tropical western Indian Ocean and the Maritime Continent exist stronger upward motions during 2006-2021, while significant downward motions are observed in the tropical eastern Indian Ocean and the tropical central Pacific (Fig. 11a). These correspond to stronger OLR negative anomalies in the tropical western Indian Ocean and the Maritime Continent, and significant OLR positive anomalies in the tropical eastern Indian Ocean and the tropical central Pacific (Fig. 11c). The interdecadal variations in the zonal circulations and OLR field are mainly influenced by the interdecadal variations in Indo-Pacific SST. During 2006–2021, significant positive SST anomalies are observed over the tropical western Indian Ocean and the Maritime Continent, while the subtropical southeastern Pacific shows pronounced negative SST anomalies (Fig. 11d). As a result, the positive SST anomalies over the tropical western Indian Ocean and the Maritime Continent induce stronger local upward motions, which in turn contribute to enhanced downward motions over the tropical eastern Indian Ocean. The significant negative SST anomalies in the subtropical southeastern Pacific enhance the zonal SST gradient, which further strengthens the Walker circulation. During 2006–2021, pronounced downward motions and positive OLR anomalies over the tropical eastern Indian Ocean are centered near 90°E, which coincides with the BSISO initiation center from day –39 to –20 (Fig. 9a, b; Fig. 10a, b; Fig. 11a–c). Therefore, the convective intensity of the BSISO at its initial stage is suppressed, which delays the onset of strong convection. In addition, during 2006–2021, the stronger upward motions over the Maritime Continent not only enhance the Walker circulation but also significantly strengthen the downward motions within 6°N–18°N via the meridional Hadley circulation (Fig. 11b). Therefore, compared with 1979–2005, the negative OLR anomalies over the northwestern Pacific dissipate more rapidly during the BSISO decay phase in 2006–2021 (Fig. 10e, f). To further verify that the interdecadal variations in the cross-month correlation feature of NWPSH are caused by the interdecadal variations in the Indo-Pacific SST climatology. Based on the summer three key SST regions in Fig. 11d, an Ind_Pac index is defined. All years are categorized into two groups based on whether the Ind_Pac index is greater than or less than zero (Table 2). In the 22 years when the Ind_Pac index exceeds zero, the correlation between June and August NWPSH is weak. Conversely, during the 21 years when the Ind_Pac index is below zero, the correlation coefficient between June and August NWPSH is 0.53, significant at the 98% confidence level (Table 3).

Fig. 11: The interdecadal variations in SST and circulation.

a Differences in summer meridionally averaged (5°S−10°N) vertical pressure velocity (W, shading, unit: ${10}^{-2}{\cdot} {\rm{Pa}}{\cdot {\rm{s}}}^{-1}$) and zonal-vertical circulation (UW, vectors, zonal wind unit: ${{\rm{m}}{\cdot} {\rm{s}}}^{-1}$, vertical pressure velocity unit: ${10}^{-2}{\cdot} {\rm{Pa}}{\cdot {\rm{s}}}^{-1}$) between 2006−2021 and 1979−2005 (2006−2021 minus 1979−2005). b is similar to (a), but for zonally averaged (100°E − 160°E) meridional-vertical circulation (VW). c is similar to (a), but for OLR spatial differences (unit: ${{\rm{W}}{\cdot} {\rm{m}}}^{-2}$). d is similar to (a), but for SST spatial differences (unit: ${\rm{K}}$). Dots indicate that the shading reaches the 95% confidence level, and only vectors reaching the 95% confidence level are displayed. The purple and black boxes in (d) denote the key SST regions for defining the Ind_Pac index.

Table 2 The years from 1979 to 2021 are classified into two categories based on whether the Ind_Pac index is greater than or less than zero

Table 3 Correlation coefficients of the NWPSH index between different summer months for positive and negative Ind_Pac index years

The Maritime Continent Weakens the Influence of the TNIO on the NWPSH after 2005

The above results explain why the BSISO displays a longer (shorter) periodicity before (after) 2005, highlighting the role of interdecadal changes in the background SST state and the resultant adjustments in the Indo-Pacific circulation. Similarly, the TNIO exerts a stronger (weaker) influence on the NWPSH before (after) 2005. The combined effects of the BSISO and TNIO result in a stronger relationship between the June and August NWPSH before 2005 than afterward. However, an issue remains unclear. Although the TNIO continues to warm from 1979 to 2024, it has a significant influence on the NWPSH only before 2005.

During 1979–2005, the TNIO warming primarily induces strong localized convergent upward motion (peaking near 60°E), which in turn generates pronounced convergent subsidence over the western Pacific, thereby strengthening the NWPSH (Fig. 12a, c, e). However, this effect vanishes after 2005. In the later period, warming over the Maritime Continent generates enhanced local upward motion, which produces stronger subsidence over the TNIO region (near 60°E). This subsidence suppresses the TNIO-induced localized upward convergence (Fig. 13). Consequently, the TNIO’s impact on the western Pacific circulation and the NWPSH weakens (Fig. 12b, d, f). The above physical mechanisms were thoroughly analyzed by us in Li et al. (2024), and further validated using SST sensitivity experiments with a dynamical model.

Fig. 12: The interdecadal variations in the impact of TNIO SST on the circulation.

a 850hPa geopotential height field (HGT, shading, unit: ${\rm{gpm}}$) and horizontal wind field (UV, vectors, unit: ${\rm{m}}{\bullet {\rm{s}}}^{-1}$) regressed onto the summer mean TNIO SST index during 1979-2005. Dots indicate that the shading reaches the 95% confidence level, and only vectors reaching the 95% confidence level are displayed. b is similar to (a), but for 2006–2024. c and d are similar to (a) and (b), but for velocity potential (VP, shading, unit: ${10}^{5}{\bullet {\rm{m}}}^{2}{\bullet {\rm{s}}}^{-1}$) and divergent wind (Div, vectors, unit: ${\rm{m}}{\bullet {\rm{s}}}^{-1}$). Only velocity potential and vectors reaching the 95% confidence level are displayed. e and f are similar to (a) and (b), but for meridionally averaged (15°S–15°N) vertical pressure velocity (W, shading, unit: ${10}^{-2}{\cdot} {\rm{P}}a{\cdot {\rm{s}}}^{-1}$) and zonal–vertical circulation (UW, vectors; zonal wind unit: ${{\rm{s}}}^{-1}$, vertical pressure velocity unit: ${10}^{-2}\bullet {\rm{P}}a{\bullet {\rm{s}}}^{-1}$). Dots indicate that the shading reaches the 95% confidence level, and only vectors reaching the 95% confidence level are displayed.

Fig. 13: The interdecadal variations in the impact of Maritime Continent SST on the circulation.

a The meridionally averaged (15°S–15°N) vertical pressure velocity (W, shading, unit: ${10}^{-2}{\cdot} {\rm{P}}a{\cdot {\rm{s}}}^{-1}$) and zonal–vertical circulation (UW, vectors; zonal wind unit: ${{\rm{s}}}^{-1}$, vertical pressure velocity unit: ${10}^{-2}{\cdot} {\rm{P}}a{\cdot {\rm{s}}}^{-1}$) regressed by the summer mean Maritime Continent SST index during 1979–2005. b is similar to (a), but for 2006–2024. Dots indicate that the shading reaches the 95% confidence level, and only vectors reaching the 95% confidence level are displayed.

Discussion

Unlike previous studies that focus solely on the NWPSH at a single timescale or examine the interdecadal variations of the summer-mean NWPSH, this paper investigates the interdecadal changes in the relationships of the NWPSH across different summer months at the intraseasonal timescale. This study identifies a unique cross-month correlation feature of the NWPSH between June and August. This phenomenon is primarily explained by the interdecadal shift in the background SST state, which leads to corresponding interdecadal variations in the BSISO periodicity. The main conclusions are as follows:

During 1979–2005, the NWPSH in June and August shows a strong and statistically significant correlation (0.55, significant at the 99.5% confidence level). However, this relationship nearly disappears during 2006–2024, with the correlation dropping to –0.05. Throughout the entire period, the correlations between adjacent summer months always remain relatively weak. The strong connection between the NWPSH in June and August also leads to the high consistencies of precipitation and circulation anomalies over the northwest Pacific and east Asia in these two months before 2005. During 1979–2005, the positive SST anomalies over the TNIO enhance subsidence and divergence over the northwestern Pacific by exciting Kelvin waves and inducing equatorial easterly anomalies, thereby strengthening the NWPSH from June to August. Meanwhile, the BSISO exhibits a more pronounced 60-day periodicity, which tends to induce opposite changes in the NWPSH between adjacent months but similar changes between June and August. Therefore, when the low-frequency, slowly varying TNIO SST forcings are superimposed with the high-frequency influences of the BSISO, they tend to weaken the relationship of NWPSH between adjacent months and strengthen the connection between the June and August.

The weakening of the cross-month correlation feature of the NWPSH after 2005 is mainly due to the significant weakening of the 60-day BSISO periodicity. The interdecadal variations in the BSISO periodicity is caused by changes in the background SST, which enhance both the zonal Walker and the meridional Hadley circulations after 2005. These changes further affect the development and decay phases of the BSISO. During 2006–2021, significant warming occurs over the Maritime Continent and the tropical western Indian Ocean, while notable cooling is observed in the subtropical southeastern Pacific. This SST anomalies pattern strengthen downward motions over the tropical eastern Indian Ocean through the Walker circulation, and enhance downward motions over the northwest Pacific via the Hadley circulation. The intensified downward motions over the tropical eastern Indian Ocean suppress BSISO convection at the initial stage and delay the onset of strong convection. The enhanced downward motions over the northwest Pacific accelerate the weakening of convection in the decay phase of the BSISO. As a result, the BSISO periodicity becomes shorter during 2006–2021.

The NWPSH exerts substantial influences on monsoon circulation, moisture transport, and rainfall distribution. Therefore, the emergence or disappearance of the June–August correlation may alter the predictability of intraseasonal and seasonal climate anomalies. In particular, during periods with a strong cross-month linkage, the NWPSH state in early summer may carry useful information for predicting its behavior in later summer, providing potential benefits for subseasonal-to-seasonal prediction.

Previous studies have shown that global warming modifies the amplitude, phase speed, and spatial location of intraseasonal oscillation, which in turn results in significant changes in extreme precipitation and temperature^41,42,43. These variations are largely driven by SST changes under a warming climate^44,45. Therefore, our investigation into the interdecadal modulation of intraseasonal oscillation periodicity induced by background SST shifts provides important insight for understanding the potential impacts of future climate change on human society.

In addition, this study only investigates the interdecadal variations in the low-frequency intraseasonal variabilities of the NWPSH, namely cross-month correlation feature. The NWPSH also exhibits a higher-frequency quasi-biweekly oscillation (QBO) in its zonal position. The QBO is primarily influenced by internal atmospheric dynamic processes, such as relative vorticity and water vapor anomalies¹². The intraseasonal precipitation variations in the Yangtze River basin are closely associated with the QBO of the NWPSH^12,46,47. However, whether the QBO of the NWPSH also exhibits interdecadal variations remains unclear and warrants further investigation.

Methods

Datasets

The monthly mean precipitation data is from the Global Precipitation Climatology Project (GPCP) dataset, covering the period from January 1979 to the present⁴⁸. The monthly and daily atmospheric circulation data is provided by the National Centers for Environmental Prediction (NCEP) and the Department of Energy Atmospheric Model Intercomparison Project reanalysis-2 (NCEP/DOE AMIP-II), including horizontal wind, vertical pressure velocity, geopotential height, relative vorticity, relative humidity, and air temperature, spanning from January 1979 to the present⁴⁹. The daily OLR data is from polar-orbiting satellites of the National Oceanic and Atmospheric Administration (NOAA), available from January 1979 to 2022⁵⁰. The monthly SST data is from the Hadley Centre Global Sea Ice and SST (HadISST) dataset, covering the period from January 1870 to the present⁵¹.

The definition methods of different indices

Based on the method proposed by Wang et al. ⁵, we first calculate the average of the maximum interannual variation center of the summer 850hPa geopotential height field over the northwest Pacific (15°N-25°N, 115°E-150°E). After standardizing this average, the result is defined as the NWPSH index, which represents the intensity of the NWPSH.

To calculate the BSISO, the daily mean OLR and 850 hPa zonal wind over the Asian summer monsoon region (10°S-40°N, 40°E-160°E) are filtered using a 20-100-day Lanczos band-pass filter with 201 weighting points^31,32. After filtering, a multivariate empirical orthogonal function (MV-EOF) analysis is performed on the OLR and 850hPa zonal wind anomalies from May 1 to October 31 to obtain the first two BSISO modes (EOF1 and EOF2) and their corresponding principal components (PC1 and PC2). Since the daily OLR data is only available up to December 2022, the effective time period after filtering is 1979–2021. Therefore, the time period for the research of BSISO in this paper is 1979–2021. To study the interdecadal variations of the intraseasonal NWPSH, the daily NWPSH index is also filtered using the 20-100-day Lanczos band-pass filter. The data from May 1 to October 31 during 1979–2021 are extracted and denoted as L-NWPSH. To investigate the impacts of BSISO on cross-month feature of NWPSH, the daily PC1 and PC2 values are averaged monthly to obtain the monthly mean $\overline{{\rm{PC}}1}$ and $\overline{{\rm{PC}}2}$.

We find that the interdecadal variations in the intraseasonal variabilities of NWPSH occur around 2005. To assess the influences of BSISO on the intraseasonal variabilities of NWPSH, composite analyses of reconstructed OLR evolution are performed for 32 and 14 BSISO events during 1979–2005 and 2006–2021, respectively. The day 0 corresponds to the date of the maximum negative anomaly of L-NWPSH and is selected according to three criteria: (1) the L-NWPSH value is less than -1.5; (2) the interval between two day 0 is at least 30 days; and (3) day 0 must falls between June 1 and August 31, when the most significant interdecadal variations in intraseasonal periodicity of NWPSH occur. Days before and after day 0 are denoted with negative and positive values, respectively (e.g., day -5 and day 5 refer to 5 days before and after day 0). The reconstructed OLR is calculated by the first modes of BSISO, the function is :

$${{OLR}}_{{Rec}}={PC}1* {{EOF}1}_{{OLR}}\,+{PC}2* {{EOF}2}_{{OLR}}$$

(1)

Specifically, ${{OLR}}_{{Rec}}$ is the reconstructed OLR field, ${{EOF}1}_{{OLR}}$ and ${{EOF}2}_{{OLR}}$ are the OLR spatial patterns of the BSISO, and PC1 and PC2 are their corresponding time series.

To assess whether the interdecadal changes in the relationship between the June and August NWPSH is significant, this study uses both the Fisher r-to-z transformation and permutation tests methods to test the interdecadal change of correlation coefficients. These two methods are widely used by meteorologists to evaluate whether changes in relationships between variables across different periods are significant^52,53,54,55.

The ENSO is represented by the Niño3.4 index, defined as the area-averaged SST over tropical Pacific (5°S–5°N, 120°W–170°W). The tropical North Indian Ocean (TNIO) index is defined as the averaged SST over the region 0–25°N, 45°E–95°E. To identify the key SST regions associated with the cross-month correlation feature of the NWPSH, an Indo-Pacific Ocean Key Region (Ind_Pac) index is constructed. The Ind_Pac index is calculated by summing the averaged SST over the tropical western Indian Ocean (15°S–10°N, 45°E–75°E) and the Maritime Continent region (15°S–10°N, 115°E–160°E), and then subtracting the SST average over the subtropical southeastern Pacific (30°S–10°S, 230°E–290°E). When using the Ind_Pac index to classify anomalous years, it is also necessary to remove the climatological mean for 1979–2021.