Weakening seasonality of Indo-Pacific warm pool size in a warming world since 1950

Seasonal variation of the Indo-Pacific warm pool (IPWP) plays an important role in oceanographic and climatological processes. While expansion of the IPWP under greenhouse warming has been widely discussed, the response of IPWP seasonality to climate change has received limited attention. In this study, we found an obvious seasonal diversity in expansion of the IPWP from 1950 to 2020, with a maximum (minimum) expansion trend of 0.28 × 107 km2/decade in winter (0.17 × 107 km2/decade in spring), which consequently reduces the seasonality amplitude of the variation in IPWP size. This is primarily attributed to the seasonal difference in the climatological spatial sea surface temperature (SST) pattern over the Indo-Pacific Ocean, especially that over the tropical Indian Ocean, which determines the capacity for IPWP expansion. Heat budget analyses show that the seasonal shortwave radiation and latent heat fluxes are the major factors controlling the capacity for change in IPWP size across seasons. The presented analyses emphasize the significant weakening of the seasonality of IPWP size, which may have great impacts on the ecological environment of the IPWP and the tropical climate system, and remind us that the intrinsic properties of the climate background of Indo-Pacific SST hold important clues about IPWP expansion under climate change.


Introduction
The Indo-Pacific warm pool (IPWP) is one of the world's major sources of heat and water vapor and affects the global climate by supporting atmospheric deep convection (Sardeshmukh andHoskins 1988, Yan et al 1992). Despite being located in the tropical ocean, where seasonality is relatively weak compared with extratropical regions, the IPWP itself experiences significant seasonal variation (Kim et al 2012, Wang et al 2019. Previous studies have revealed that the sea surface temperature (SST) and IPWP size are characterized by a noticeable semiannual cycle that is influenced by solar radiation and the Indian summer monsoon. The warmest SST and maximum size of the IPWP are observed in boreal spring, while the coolest SST and minimum size occur in boreal winter (Fasullo and Webster 1999, Kim et al 2012, Wang et al 2019. Further, the Indian Ocean warm pool (IOWP) has a stronger seasonal variability than the western Pacific Ocean warm pool (WPWP) (Kim et al 2012, Zhang et al 2016, Wang et al 2019. In addition to its characteristics, the impacts of the IPWP on climate also exhibit obvious seasonality. For example, the variability of winter IPWP is a major factor controlling the Madden-Julian Oscillation (MJO; Roxy et al 2019); that of spring IPWP affects the summer monsoon in East Asia and the South China Sea (Yin et al 2020) and the precipitation in Africa through its affect on Walker circulation (Williams and Funk 2011); that of summer IPWP is crucial to the genesis and development of tropical cyclones in the Indo-Pacific region (Yang and Zheng 2018); and that of autumn IPWP could affect precipitation in western China (Wei et al 2018), etc.
Because of the important role of the IPWP in the global climate system, expansion of the IPWP under global warming has been a topic of interest for years (Weller et al 2016, Sipala et al 2019, Park et al 2022. A recent publication reported that the size of the IPWP doubled from 1900-1980to 1981(Roxy et al 2019. Climate model simulations showed that 85% of the IPWP's expansion in the past decades is a result of human influence (Bai et al 2022). The influence of the recent human-caused expansion of the IPWP on climate systems has been widely discussed, for example Hadley circulation, Walker circulation (Brierley et al 2009, Williams and Funk 2011, tropical cyclones (Webster et al 2006, Benestad 2009), East Asian summer monsoon , Yin et al 2020, Jian et al 2022, MJO (Roxy et al 2019, Dasgupta et al 2020, the El Niño-Southern Oscillation (ENSO) (Sun 2003) and precipitation over different regions (Zhou 2014). Besides, expansion of the IPWP will also have important impacts on its ecological environment, such as changing the species composition and improving the productivity of mangroves in Indonesia (Field 1995, Shearman et al 2013, Patrick et al 2016, changing the life cycle of planktic organisms in the IPWP (Patrick 2016), promoting the growth rate and biomass accumulation of phytoplankton (Winder and Sommer 2012) and altering the patterns and predictors of marine biodiversity (Tittensor et al 2010).
Although expansion of the IPWP under greenhouse warming has been widely discussed, the response of IPWP seasonality to climate change has received limited attention. According to several previous independent studies, we notice that the changes in IPWP size may vary across seasons. For example, while the annual size of the IPWP increased by about 30% from 1953 to 2012 (Weller et al 2016, Bai et al 2022, winter expansion of IPWP was larger than 60% during the same period (figure 1(c) of Roxy et al 2019). This implies that the IPWP expansion trend may vary in different seasons and suggests that the seasonality of the IPWP may change under global warming. Thus, this study aims to study the changes in IPWP seasonality under the past greenhouse warming background and the main factors causing these changes.

Data and methods
The analyses in this work were carried out based on the following datasets: (a) the Met Office Hadley Center Sea Surface Temperature version 1 (HadIS-STv1) dataset (Rayner et al 2003), (b) the National Oceanic and Atmospheric Administration Extended Reconstructed Sea Surface Temperature version 5 (ERSSTv5) dataset (Huang et al 2017), (c) the Institute of Atmospheric Physics (IAP) ocean temperature dataset (Cheng et al 2017), (d) the National Centers of Environmental Prediction Global Ocean Data Assimilation System (NCEP-GODAS) ocean current dataset (Behringer et al 1998) and (e) the National Centers for Environmental Prediction-Department of Energy (NCEP-DOE) reanalysis 2 surface heat fluxes data, including shortwave radiation (SWR), longwave radiation (LWR), latent heat flux (LHF) and sensible heat flux (SHF) (Kanamitsu et al 2002).
The IPWP size is defined as the area where SSTs are greater than 28 • C in the tropical Indo-Pacific region (25 • S-30 • N, 40 • E-140 • W). In this study, the IPWP is divided into the Indian and Pacific Ocean sectors (see figure S1), respectively referred to as the IOWP and the WPWP, based on the basin mask information from NOAA's National Oceanographic Data Center (Locarnini et al 2010). The image histogram approach (Chen et al 2002, Cheng and Shi 2004, Xu et al 2010, Wong et al 2016, Leung et al 2022 was applied to display the spatial probability frequency distributions (PFDs) of SST and to estimate the capacity for change in warm pool size under ocean warming. Heat budget analyses were used to diagnose the seasonal variation of climatological SST. More details about the calculations and explanations about the methodology can be found in the supplementary materials. All linear trend analysis results presented below are statistically significant at the 99% confidence level, unless specified otherwise.

Weakening seasonality of the IPWP size since 1950
The long-term changes in IPWP seasonality could provide important clues about how anthropogenic climate change might affect the global climate through the IPWP. Here, we first present a comparative analysis of the size of the IPWP and its long-term change in different seasons from 1950-2020, which shows an obvious seasonal diversity of the expansion trends of the IPWP.
The long-term changes in IPWP size differ greatly between seasons, with the largest expansion trend in winter [December-January-February (DJF), 0.28 × 10 7 km 2 /decade], followed by autumn [September-October-November (SON), 0.26 × 10 7 km 2 /decade]. The increases in IPWP size are relatively smaller in summer [June-July-August (JJA), 0.18 × 10 7 km 2 /decade] and spring [March-April-May (MAM), 0.17 × 10 7 km 2 /decade] (figure 1(a)). It is worth noting that the rising trend of IPWP size in winter is almost twice as large as that in spring and summer. The seasonal differences in IPWP expansion exist not only in the long-term trends but also in the spatial patterns. We find that the winter and autumn IPWP expand greatly to the west in the Indian Ocean sector in addition to the apparent eastward expansion in the Pacific sector (figures 1(c) and (f)), but in the spring and summer the IPWP mainly expands eastward in the Pacific sector (figures 1(d) and (e)). The different expansion patterns in Indian Ocean sectors cause diverse forcing effects on the Walker circulation, with the fast expanding winter and autumn IPWP in the Indian Ocean sector coupling with the reinforced and westward-shifted Walker circulation in the Indian ocean sector (figure S2). Interestingly, despite the great differences in IPWP size changes in different seasons, the increasing trends of the Indo-Pacific SST are rather constant across seasons (∼0.09 • C/decade; figure 1(b)) (Xie et al 2010, Huang et al 2013, Geng et al 2020. These results imply a significant seasonal diversity in the IPWP expansion trend under the seasonally homogeneous ocean warming background. The large seasonal diversity in the IPWP expansion rate indicates that the seasonality of the IPWP has been changing in the past few decades. By dividing the period 1950-2020 into two halves, figure 2(a) compares the annual cycle of the IPWP in the pre-1985 period (P1, 1950-1984) and the post-1985 period (P2, 1985-2020). The seasonal cycle of IPWP size, with a bimodal pattern peaking in spring and autumn, has generally shifted upward under the background of global warming. In particular, the expansion trends in winter and autumn (secondary peak) were larger than those in spring and summer (primary peak) (bars in figure 2(a)). This non-uniform change made the secondary peak of the seasonal cycle rise to a larger extent compared with the primary peak. As a result, a decrease in the amplitude of the seasonal cycle of IPWP size is observed in P2 relative to P1. Quantitatively, the seasonal cycle amplitude (defined as Peak−Trough Average × 100%) reduces from 43% to 29% from P1 to P2. The decrease in the seasonal cycle amplitude can also be clearly demonstrated from the 4-15 months bandpass filtered IPWP size series, which mainly captures the signal of the seasonal cycle (figure 2(b)), with its annual standard deviation decreasing at a rate of −4.0 × 10 5 km 2 /decade (figure 2(c)). Similar conclusions can be obtained from observations from 1979 to 2020 (figures S3(a)-(d)) and those based on the ERSSTv5 dataset (figures S4 and S5). Besides, the weakening seasonality is observed not only for the size of the IPWP but also for its volume (figures S6(a) and (b)). An interesting phenomenon is that there seems to be a reverse change in the seasonal cycle amplitude of the shallow IPWP and deep IPWP size (figures S6(c)-(f)).

Contributions of the IOWP and WPWP to the weakening seasonality of the IPWP
According to previous research, the variabilities and characteristics of the IOWP differ largely from those of the WPWP. Schneider et al (1996) concluded that SWR affects the vertical temperature structure more in the Pacific than in the Indian Ocean. Fasullo and Webster (1999) found that the IOWP SST is less sensitive to surface thermal forcing than that in the Pacific, and is influenced by the Indian summer monsoon (Kim et al 2012, Wang et al 2019. Thus, we further divide the IPWP into the IOWP and WPWP (figure S1), and examine their contributions to the weakening IPWP seasonality. Surprisingly, we find that the seasonal diversity of the IPWP expansion trend is almost solely contributed by the IOWP.
As shown in figures 2(d)-(i), the IOWP and WPWP exhibit different seasonality, although both are primarily dominated by a 12-month cycle. For the IOWP, the peak (trough) of its seasonal cycle is in spring (late summer), with a large seasonal cycle amplitude (104%) in the period 1950-2020 ( figure 2(d)). In contrast, the WPWP seasonal cycle has its peak (trough) in autumn (winter) and its amplitude (36%) is much smaller than that of the IOWP ( figure 2(g)). This result indicates a stronger fluctuation of the IOWP seasonal cycle than that of the WPWP, which is consistent with previous findings (Kim et al 2012, Wang et al 2019. Similar to the IPWP, the change in size of the IOWP from P1 to P2 also exhibits obvious seasonal diversity, showing that the largest expansion trend in winter (0.17 × 10 7 km 2 /decade) is more than twice as large as that in spring (0.07 × 10 7 km 2 /decade) ( figure 2(d)). This indicates a long-term change in IOWP seasonality. Because the speed of IOWP expansion in winter and autumn is twice that in spring and summer (figure 2(d)), the seasonal variation in IOWP size has been diminished since 1950, with its seasonal cycle amplitude dropping from 127% to 88% from P1 to P2. This is also evident in the 4-15 months bandpass filtered IOWP size series and its annual standard deviation, which depicts a downward trend of −2.0 × 10 5 km 2 /decade (figures 2(e) and (f)). These changing patterns in IOWP size are found to be consistent with those of the IPWP.
On the contrary, although the variation in WPWP size does exhibit obvious seasonality, the WPWP expands almost identically in all seasons from P1 to P2, with its size increase of each month ranging from 0.09 × 10 7 to 0.13 × 10 7 km 2 (figure 2(g)). Consequently, the seasonal cycle of WPWP size shifts upward approximately uniformly (figure 2(g)). The seasonal cycle amplitude stays on almost the same level (38% in P1 and 34% in P2), and the standard deviation of 4-15 months band-pass filtered WPWP size shows a statistically insignificant long-term trend (−4.0 × 10 4 km 2 /decade; figures 2(h) and (i)). Based on the comparison of the changes in the seasonality of the IOWP and WPWP sizes (figures 2(d) and (g)) as well as the spatial patterns of IPWP expansion in four seasons (figures 1(c)-(f)), we conclude that the seasonal difference in IOWP expansion weakens the amplitude of IPWP seasonality. Consistent results are also obtained from ERSSTv5 (figure S7).

Causes of the seasonal diversity of IPWP expansion under tropical warming
The seasonal diversity of IPWP expansion somehow contradicts the seasonally homogeneous SST warming over the Indo-Pacific region. The overall regionalmean SST trend ( figure 1(b)) and the spatial pattern of SST change show no obvious seasonal differences (figures S8 and S9). In this section, based on the image histogram approach (see supplementary materials), we show that the seasonally diverse IPWP expansion speed is mainly due to the seasonality of the climatological Indo-Pacific SST pattern.
Comparing periods P1 and P2, the SST PFD of each season (left column of figures 3(a)-(d)) shifts to the right, to the first order, without obvious changes in shape and skewness (figure 3(e)), which increases the area where SST ⩾ 28 • C. The change in IPWP size largely depends on the PFD shape. For example, in P1, the Indo-Pacific basin SST is relatively cool and the PFD peak is near 28 • C in winter, indicating a large number of grids (area) whose SSTs were slightly lower than 28 • C ( figure 3(a)). These grids will be identified as part of the IPWP in P2 as the ocean warms. On the other hand, there are relatively fewer grids (area) whose SSTs are slightly lower than 28 • C in spring ( figure 3(b)). Hence, the change in IPWP size in spring is significantly smaller, although the increase in SST is rather similar in spring and winter (figures 1(b), S8 and S9). In other words, the SST PFD shape, which is associated with the climatological SST pattern, indicates the potential for IPWP expansion, namely the capacity for change in IPWP size (Kajtar et al 2021, Park et al 2022. The seasonal difference in the capacity for IPWP size change may be the main cause of the weakening of IPWP seasonality in the past decades. Two ideal experiments (see supplementary materials) which assume (a) seasonal uniform warming only (SUE) and (b) seasonally and spatially homogeneous warming (SSUE) were further carried out. Results show that both SUE (purple line) and SSUE (green line) generally capture the observed PFD changes of different seasons (figures 3(a)-(d)), although the SST PFD shapes under SSUE are slightly different from reality because the observed spatially uneven warming pattern is neglected. The amplitude of the IPWP seasonal cycle under SSUE (SUE) drops from 43% to 28% (43% to 29%) from P1 to P2, being consistent with the observed change (14%). The IPWP size changes under both SSUE (1.02 × 10 7 , 0.67 × 10 7 , 0.70 × 10 7 and 0.86 × 10 7 km 2 in winter, spring, summer and autumn, respectively) and SUE (1.10 × 10 7 , 0.66 × 10 7 , 0.68 × 10 7 and 0.86 × 10 7 , respectively), consistent with the observed changes (1.03 × 10 7 , 0.60 × 10 7 , 0.65 × 10 7 and 0.89 × 10 7 , respectively) ( figure 3(f)).
Besides the small differences [(0.01-0.07) × 10 7 km 2 ] in IPWP size change between SSUE (i.e. the capacity for IPWP size change) and observation, consistent results are also evident in the seasonal diversity ( figure 3(f)). In addition, the spatial patterns of the capacity for IPWP change and the IPWP change in SUE from P1 to P2 are also consistent with observation (figures 3(g)-(j)). The presented ideal experiments prove that the seasonality of the capacity for IPWP size change plays a dominant role in changing the seasonal variability of the IPWP, and emphasizes the importance of the intrinsic properties of the climatological SST pattern over the Indo-Pacific Ocean in analyzing seasonal IPWP expansion. Meanwhile, the contribution of spatially inhomogeneous SST change is fairly small when assessing the seasonality of the capacity for IPWP size change.
The ideal experiments further confirm that that the seasonality change in IPWP size is mainly determined by the capacity for IOWP size change (0.60 × 10 7 , 0.27 × 10 7 , 0.22× 10 7 and 0.39 × 10 7 km 2 in winter, spring, summer and autumn, respectively) ( figure 3(f)). The seasonal cycle amplitude of the capacity for IOWP size change decreases from 127% to 89%, which is highly consistent with observations (127% to 88%) and consistent with that of the IPWP. On the contrary, the capacity for WPWP size change does not show obvious seasonal differences (figure 3(f)). Consequently, the amplitude of the seasonal cycle of WPWP size shows no significant change (from 38% to 33%).
The above analysis emphasizes the crucial impacts of seasonality in the climatological Indian Ocean SST pattern that determines the capacity for IOWP expansion in weakening the IPWP seasonality. Based on heat budget analyses (see supplementary materials), we further show that that the SWR and LHF primarily determine the climatological SST pattern, with LHF playing a key role in the tropical Indian Ocean because of the strong Indian winter and summer monsoons (figures 4(a) and (b)). In winter in the Bay of Bengal and Arabian Sea, the negative LHF anomaly caused by the strong Indian winter monsoon and negative SWR anomaly leads to a negative net surface heat flux anomaly, which further results in a negative anomaly in SST tendency (figures 4(c), (d), (f) and (g)). The warm advection in the Arabian Sea weakens the negative anomaly in SST tendency to some extent ( figure 4(e)). The negative anomaly in SST tendency couples with the cool SST pattern over the tropical Indian Ocean and may provide a large capacity for IPWP expansion in winter. In the tropical Indian Ocean in spring, the positive LHF and SWR anomalies lead to a positive net surface heat flux  anomaly (figures 4(c), (d), (f) and (g)), which results in the warm SST pattern over the tropical Indian Ocean and a smaller capacity for IPWP expansion. The heat budget analysis reveals that it is the seasonal differences in atmospheric forcing (especially SWR and LHF) that determine the seasonality of the climatological SST pattern over the tropical Indian Ocean, which further influences the seasonal diversity of IPWP expansion.

Conclusions and discussion
In this study we have revealed the changes in the seasonality of the size of the IPWP under recent global warming. Obvious seasonal diversity is observed in expansion of the IPWP since 1950. The rate of expansion of the IPWP in winter is the largest and almost twice that in spring. The winter and autumn IPWP expand not only in the Pacific section but also in the Indian Ocean section, while the spring and summer IPWP show no obvious expansion in the Indian Ocean section. The large seasonal diversity in IPWP expansion rate reduces the seasonality of IPWP size. Based on two ideal experiments, we confirm that the seasonality of the capacity for IPWP size change, which is determined by the seasonal climatological SST pattern, is the main cause of the weakening IPWP seasonality. In particular, the seasonality of the Indian Ocean SST pattern, which is mainly determined by the seasonal SWR and LHFs, is shown to be the major contributor to the weakening amplitude of the IPWP seasonal cycle.
Previous studies have shown that climate change is reflected not only in changes in the mean values of climate variables but also in the changes in their seasonal cycles (Wu et al 2008, Qian et al 2011, Stine and Huybers 2012, Qian and Zhang 2019. In particular, Qian et al (2011) revealed that the change in seasonal cycle could affect trend calculations because the deseasonalizaton method used removed the repetitive climatological annual cycle. Therefore, the decreasing amplitude of the seasonal cycle of IPWP size may affect the estimations of IPWP size anomalies and long-term trends in the context of global warming.
In earlier work, Park et al (2022) found that climate models which simulate a smaller climatological IPWP size would project a larger IPWP expansion in the future, and vice versa, which indicates that the projected IPWP expansion speeds are associated with the simulated intrinsic properties of the IPWP and tropical SST pattern in climate models. Also, Geng et al (2020) concluded that the seasonal change in tropical precipitation is dominated by the seasonal cycle of background tropical SST rather than the seasonal diversity of SST change, which shows only a small magnitude. Meanwhile, in this study, we conclude that the changes in IPWP seasonality in the past decades are related to the seasonal difference in climatological Indo-Pacific SST patterns. The above analyses demonstrate that the seasonal diversity in the climate SST background does have an important impact on the response of IPWP seasonality to climate change, which may have been underestimated previously. The large seasonal diversity in IPWP expansion will continue in the near further and this may have great impacts on the tropical climate, such as the location and intensity of Walker circulation and the ecological environment of the IPWP in different seasons, which deserve close scrutiny and further study.