3.1. Diversity of WPI in CMIP6 Models
The WPI is defined by the averaged SST in the warm pool region (15º S–15º N, 90º E–150º E) relative to the tropical mean SST (15º S–15º N, 0º–360º E) in the PD climate. We also tested the sensitivity of the definition by shrinking/enlarging the domain and found consistent results. As shown in Fig. 1, the climatological warm pool in the PD climate is warmer than the tropical mean SST; thus, the warm pool index is positive in all climate models. The multimodel ensemble (MME) mean (1.05°C) of the 39 models is comparable to the observed WPI. However, the WPI shows a larger intermodel spread ranging from 0.53°C to 1.50°C. Compared with the MME mean, 17 of the 39 models show a weaker WPI, whereas 22 of the 39 models show a stronger WPI. To investigate how the PD WPI is linked to the future precipitation changes in the tropics, two groups have been selected (10 models each) based on the strength of WPI: the strong warm pool group (SWG) and the weak warm pool group (WWG; Fig. 1 and Table S1). The difference between the SWG (1.36°C) and WWG (0.71°C) remains significant difference (0.65°C), showing the uncertainty in the simulated WPI in the PD climate. Instead of uncertainties in intensity, previous studies mainly focused on uncertainties associated with the morphology (i.e., area) of the IPWP in climate models (i.e., Yang et al. 2020; Park et al. 2022a, b). In this study, we show that WPI in the PD climate is diverse and may contribute to profound changes in the future climate projections of climate models.
To examine the impact of WPI on the differences in spatial patterns between the two groups, we first analyzed the patterns of SST, precipitation, and low-level winds in the PD climate (Fig. 2). The SST near the Maritime Continent (MC) region in the SWG is warmer by definition, and the low-level convergence is stronger than that in the WWG (Fig. 2a). Studies have shown that the IPWP serves as an atmospheric “water tower” because of warmer SSTs, relatively low surface pressure, and strong moisture convergence (Jian et al. 2022). Consistent with the previous findings, the difference in precipitation exhibits enhanced wetter conditions over the MC region in the SWG (Fig. 2b). In addition, the central Pacific (CP) region, which displays slightly cooler SSTs in SWG models (Fig. 2a), exhibits a weakening of climatological precipitation (Fig. 2b). Thus, in addition to the differences in WPI and low-level circulation, the two model groups display a robust difference in the precipitation between the MC and CP regions in the PD climate. As the model biases and intermodel variability remain major challenges for reducing the model uncertainty of future projections (Zheng 2019), it is important to understand how diversity in WPI and associated tropical convection in the PD climate influence future changes in the tropical air–sea coupled system under greenhouse warming.
3.2. Diversity of projected SST and precipitation patterns in CMIP6 Models
To show the differences in the spatial patterns of the CMIP6 models under anthropogenic global warming, we display the change (SSP5-8.5 minus historical) in SST warming, low-level circulation, and precipitation for the SWG (Fig. 3a) and WWG (Fig. 3b) models. Consistent with previous findings (Zhang et al. 2019; Fosu et al. 2020), both the SWG and WWG models indicate strong and significant SST warming in the equatorial Pacific with dominant westerlies over the equatorial Pacific, suggesting El Nino-like warming. Over the Indian Ocean, there is strong warming in the western Indian Ocean with dominant easterlies over the equatorial Indian Ocean. In addition, precipitation increases over the regions with warmer SSTs in both groups. However, there are distinctive differences in the projected SST warming and precipitation patterns between the two groups. For example, the change in precipitation shows a significant difference in the CP region. Thus, the SWG models exhibit relatively strong SST warming, low-level circulation, and precipitation compared with the WWG models, which intimate the potential impact of WPI in the PD climate.
Now, we examine the difference in ∆SST (SSP minus historical) and ∆precipitation between the two model groups (Fig. 4). The models with the intense warm pool in the PD climate tend to simulate enhanced (suppressed) wet (dry) conditions over the CP (MC) region, west of the dateline under greenhouse warming (Fig. 4a). We also evaluated the changes in ∆precipitation in the CP and MC regions from both groups and found consistent and significant results (Table 1). Although precipitation changes in the CP region are strong and positive in both model groups, precipitation changes remain weakly positive over the MC region in the WWG models. In addition, the difference in the precipitation change pattern shown in Fig. 4a is the opposite of the difference in the precipitation pattern in the historical simulation shown in Fig. 2b. In addition, the low-level convergence that appeared over the MC region in the SWG models in the PD climate is replaced by low-level divergence over the same region, because of the eastward shift of the convection from the MC region under greenhouse warming.
Table 1
∆Precipitation (mm/day) of MC and CP regions and their statistical significance.
| MC | CP | T-test significance |
SWG | -0.48 ± 0.65 | 2.17 ± 1.13 | P < 0.01 |
WWG | 0.25 ± 0.37 | 0.86 ± 0.74 | P = 0.03 |
T-test significance | P < 0.01 | P < 0.01 | |
Although all models displayed the strongest SST warming in the equatorial eastern Pacific under anthropogenic GHG forcing, the most significant difference in ∆SST between the two groups emerge around the dateline (180º E, Fig. 4b). This means that despite consistently strongest SST warming in the eastern equatorial Pacific, the models with the intense warm pool in the PD climate also simulate stronger and more significant SST warming in the CP region than the models with weaker WPI (Fig. 4b). In this regard, the pattern of the ∆SST difference is consistent with that of the ∆precipitation difference between the two groups (Fig. 4). The shift of the convective zone to the CP region weakens the ascending branch of the Walker circulation over the MC region, which in turn weakens the easterlies in the equatorial Pacific, resulting in anomalous westerlies over the western-central Pacific. Therefore, the robust SST warming displayed in the SWG models under anthropogenic GHG warming is associated with the precipitation change in the MC and CP regions. Thus, our results suggest that the diverse SST warming pattern depicted in the CMIP6 models probably resulted from the precipitation pattern modulated by the intensity of the warm pool in the PD climate.
Therefore, based on the significant differences displayed in Fig. 4, we suggest that the diversity of WPI in the PD climate can produce significant changes in the precipitation over MC (enhanced drier conditions) and CP (enhanced wetter conditions) under greenhouse warming. Such distinct changes in the projected precipitation between the MC and CP regions will enforce low-level westerly anomalies over the western-central Pacific, which in turn facilitates the relative SST warming in the CP region. In turn, the CP SST warming may strengthen the El Niño-like warming in the equatorial Pacific. However, in this study, we mainly focus on the WPI-induced precipitation pattern change.
To further support our argument, we checked individual models and their intermodel relationship in Fig. 5 (spatial correlation patterns are given in Fig. S1). Specific information about the box regions used to calculate the correlations is given in Table S2. The WPI shows a significantly strong correlation with the difference (CP minus MC) of ∆precipitation (Fig. 5a and Figure S1a), indicating its significant impact on the projected precipitation pattern. Despite the strong correlations, there is a notable feature between the SWG and WWG groups. While the WWG models consistently exhibit weaker precipitation differences between the CP and MC regions, the SWG models exhibit significant diversity. This suggests that the precipitation response may be affected by other factors when the climatological warm pool SST is sufficiently high. Notably, the observed WPI is located between the SWG and WWG, and it is slightly closer to the mean of the SWG group. Thus, it is higher than the MME mean of the whole model. According to the emergent constraint with the significant correlation between WPI and ∆precipitation, we may expect that future changes in precipitation will be wetter in the CP region and drier in the MC region than those of the current models' projections.
The differences in the precipitation change can be associated with the differences in the strength of the circulation. Therefore, the WPI has a relatively strong and significant correlation with the western Pacific zonal wind change (∆U), suggesting that stronger climatological WPI leads to anomalous westerlies in the equatorial western Pacific under greenhouse warming (Fig. 5b and Fig. S1b). Consistently, the WPI exhibits a positive relationship with CP warming even though the correlation is relatively weak (Fig. 5c). The correlation exceeds 0.5 in some regions, west of the dateline (Fig. S1d).
The strong relations of the tropical oceanic and atmospheric responses with the WPI are due to strong oceanic–atmospheric interactions in these regions. As depicted in Fig. 6a, a stronger correlation exists between ∆precipitation and the western Pacific ∆U. This suggests that the persistent westerlies over the western Pacific can lead to enhanced precipitation over the CP region, which promotes low-level westerlies again. The anomalous westerlies induce oceanic Kelvin waves and eastward surface currents, which induce CP warming; thus, their relationship is shown in Fig. 6b. Hence, models with an intense warm pool in the PD climate may lead to CP SST warming under anthropogenic GHG forcing through the modulation of MC and CP precipitation, followed by the surface wind responses in the western Pacific (Fig. 6c). In addition, this strong and significant correlation between ∆CP SST and ∆tropical Pacific SST relative to the tropical warming indicates the potential impact of relative warming in the CP region on the tropical Pacific SST warming (Fig. 6d and Fig. S1e).
Thus far, we have shown that the climatological WPI is closely related to precipitation, wind, and SST responses to greenhouse forcing, and this strong relationship is accompanied by strong ocean–atmosphere interactions. However, the scientific question of how the WPI drives the ocean–atmosphere interactions under greenhouse warming remains unanswered. To address this scientific question, we analyzed the precipitation sensitivity in the MC and CP regions in the historical simulation and examined its relationship with the WPI (Fig. 7a). Here precipitation sensitivity is defined considering the difference in precipitation (CP minus MC) and the SST (10º S–10º N, 95º E–180º E) on an interannual timescale. This sensitivity measures how tropical precipitation responds to warming in the warm pool region on interannual time scales and will provide a clue on how to respond to long-term external forcing such as greenhouse warming. Precipitation sensitivity tends to increase as the WPI increases. Although there is strong diversity among the SWG models, the WWG models show weaker sensitivity than the SWG models (Fig. 7a). Despite the differences in precipitation sensitivity between the MC and CP regions, the SWG models still display enhanced (weakened) precipitation climatology over the MC (CP) region compared with the WWG models (Fig. 2b). The correlation between the warm pool index and sensitivity is 0.43, which is significant at the 99% confidence level. If the warm pool SST is sufficiently warm compared with the other tropical oceans, additional warming may not be critical for anomalous convection, and other factors will determine the convective activity (Yun et al. 2021). However, the surrounding ocean of the warm pool, where SST is lower than the warm pool, may be sensitive to the warming, allowing for enhanced convection there. In particular, convection can be easily enhanced in the equatorial western-CP, where the horizontal SST gradient is strong. On the other hand, convection within the warm pool can be suppressed in response to the enhanced convection in the surrounding area, which prefers the dipole precipitation pattern.
This mechanism might work in the response to long-term climate changes. Greenhouse warming enhances the overall convection in the tropics, but the enhanced convection can be stronger in the surrounding area of the warm pool when the climatological warm pool is stronger. The enhanced convection results in suppressed convection, allowing us to calculate the correlations between the precipitation sensitivity in the PD climate and ∆precipitation in the tropics (Fig. 7b). Results showed that significant positive correlations appear over the western-CP, south Asia, and the central Indian Ocean, and negative correlations appear in the eastern Indian Ocean, Australia, and MC. It is striking that the pattern in Fig. 7b is similar to the precipitation difference between the SWG and WWG models depicted in Fig. 4a, implying that WPI-related convection sensitivity determines future changes in the precipitation in the Indo-Pacific Oceans.