1. Introduction
The tropical Indian Ocean (TIO) is an area of active air–sea interactions, due to the background Indo-Pacific warm pool and the absence of steady equatorial wind and upwelling systems (Li et al. 2003; Cai et al. 2013). The Indian Ocean dipole (IOD) is the most remarkable interannual phenomenon in the TIO (Saji et al. 1999; Webster et al. 1999), which causes weather or climate anomalies not only in the nearby Indian Ocean region but also worldwide (Guan and Yamagata 2003; Annamalai et al. 2003; Xu et al. 2020). The positive phase of the IOD (pIOD) is associated with increased precipitation over East Africa and drought over the Maritime Continent (Endo and Tozuka 2016), bushfires in the Australian mainland (Cai et al. 2009), and more autumn precipitation in South China (Qiu et al. 2014). Research into IOD mechanisms can be added to the evaluation of physical processes in models, ultimately benefiting the forecasting and prediction of weather and climate related to the IOD.
The IOD is characterized by a zonal sea surface temperature (SST) anomaly (SSTA) gradient in the TIO (Tozuka et al. 2008). During pIOD, the eastern TIO (ETIO) becomes anomalously cooler while the western TIO (WTIO) becomes anomalously warmer, constituting a dipole pattern (Saji et al. 1999). IOD events typically occur in the boreal summer and fall, with the autumn season being regarded as the mature phase of the IOD (Saji et al. 1999; Du et al. 2013). The phase locking of the IOD is intimately linked to the Indian Ocean summer monsoon variability. Generally, stronger southeasterly trade winds along the Sumatra–Java coast induce local upwelling, leading to anomalous cooling and supporting the development of pIOD (Luo et al. 2008), which is the fundamental development mechanism for the IOD (referred to as local Bjerknes feedback; Bjerknes 1969). In boreal December, pIOD events abruptly decay as the winds change to westerlies, marking the onset of the Australian summer monsoon (Jourdain et al. 2013).
Positive IOD skewness is the tendency for pIOD events to have a larger amplitude than negative IOD (nIOD) events and to be more extreme than nIOD (Cai and Qiu 2013). Interannual anomalies of SST, wind, and precipitation over the southern ETIO (SETIO) all exhibit negative skewness (Ogata et al. 2013). Extreme pIOD is the main source of positive IOD skewness (Hong et al. 2008a). The observed extreme IOD events, which happened in 1994, 1997, and 2019, brought vast hazards to nature and society (Page et al. 2002; Xu et al. 2020) but also created a burst of scientific activity. For example, the 1997 extreme pIOD event directly led to the birth of the IOD concept (Webster et al. 1999; Saji et al. 1999).
However, the cause of the IOD skewness is under debate. One of the fundamental reasons for the IOD asymmetry is related to the mean-state deep thermocline in the ETIO, since it leads to minimal surface warming in response to a deeper thermocline than a shallower thermocline perturbed by winds (Cai and Qiu 2013). Based on sensitivity experiments with an ocean general circulation model, Ogata et al. (2013) showed that negative SST skewness in the ETIO emerges in response to symmetric zonal wind fluctuations, indicating the important role of the mean-state deep thermocline. Another important mechanism is the nonlinear dynamic heating (NDH) process (Hong et al. 2008a,b). Hong et al. (2008b) pointed out that IOD skewness is attributed to the asymmetry of the wind stress–ocean advection–SST feedback, in which the nonlinear advection tends to cool the eastern upper ocean. Yang et al. (2020) also emphasized the role of the NDH in the nonlinearity of pIOD events. However, Hong and Li (2010) found that the ratio of the mixed‐layer temperature (MLT) tendency between the positive and negative events remains approximately unchanged with and without the vertical temperature advection. Instead, the sea surface height (SSH)–SST asymmetry is primarily attributed to the asymmetric atmospheric heating and wind responses to the ETIO SSTAs (Hong and Li 2010).
Asymmetric SST–cloud–radiation (SCR) feedback is one of the atmospheric effects that may contribute to the positive IOD skewness (Hong et al. 2008a). This feedback implies that warm SSTAs during nIOD are thermally damped by negative SCR feedback, while cold SSTAs during pIOD are not affected by SCR feedback under clear skies and can intensify faster (Hong et al. 2008a). However, Cai et al. (2012) argued that the SCR feedback increases with the development of cold ETIO SSTAs and the IOD skewness does not occur because of a breakdown in the SCR feedback. In addition, the IOD asymmetry can be enhanced by a nonlinear barrier layer response, which shows a greater barrier layer thinning to colder ETIO and less barrier layer thickening to warmer ETIO (Masson et al. 2004; Qiu et al. 2012; Cai and Qiu 2013). Nakazato et al. (2021) found that vertical mixing also plays an important role in causing IOD asymmetry in SST using a regional ocean model. Due to the Indo-Pacific interactions (Cai et al. 2019), IOD skewness is also modulated by El Niño–Southern Oscillation (ENSO) (Schott et al. 2009; Xie et al. 2010; Abram et al. 2020; An et al. 2023).
Coupled general circulation models participating in the Coupled Model Intercomparison Project (CMIP; Meehl et al. 2007; Taylor et al. 2012; Eyring et al. 2016) are effective tools for understanding the underlying physical IOD mechanisms and feedbacks, and their reactions to greenhouse warming. Ng and Cai (2016) revealed that CMIP5 models produce overly weak present-day westerlies in the mean state due to the positive IOD-like biases (Li et al. 2015; Long et al. 2020) and then generate weak present-day IOD skewness in the historical runs. Previous studies also pointed out that the IOD skewness is projected to weaken in a warm climate due to the enhanced SST response to thermocline changes in the CMIP5 models (Zheng et al. 2013; Cai et al. 2013; Ng et al. 2014), which is driven by the faster warming in the WTIO than ETIO and the slowdown of the Indian Walker circulation (Vecchi et al. 2006; Vecchi and Soden 2007). This nonuniform TIO warming under warming climate may induce larger nonlinear zonal temperature advection, leading to more frequent occurrences of extreme pIOD events (Cai et al. 2014, 2021). Moreover, responses of the zonal wind and current to IOD are also related to the simulation of atmospheric stability: increased stability results in reduced variance in the zonal wind, ocean current, and even thermocline, possibly contributing to a reduction in NDH (Cai et al. 2013; Zheng et al. 2013).
Understanding how mean-state biases affect simulated IOD skewness may improve climate projections in IOD-affected countries. Although previous studies have pointed out some reasons for the biases in simulated IOD skewness, the exact mechanisms of underestimated IOD skewness remain uncertain. Additionally, previous findings are mostly based on CMIP5 models and the performance of IOD skewness in CMIP6 models needs further research. This study’s primary objective is to evaluate whether CMIP6 models demonstrate improvements in replicating IOD skewness compared to CMIP5 models. While McKenna et al. (2020) previously assessed the performance of IOD skewness in CMIP5 and CMIP6 models, our study focuses more on the origins behind the underestimation of IOD skewness. We will explore the specific processes influencing IOD skewness simulation in CMIP6 models compared to CMIP5 models, contributing to a comprehensive understanding of this crucial climate variable.
We investigate the IOD skewness simulation from CMIP5 to CMIP6 as well as the effects of mean-state biases. The paper is organized as follows. Section 2 introduces the data and methods. Section 3 shows the performance of CMIP5/6 models in simulating the IOD skewness and some dynamical processes related to the IOD asymmetry. In section 4, we examine the simulated nonlinear zonal advection in more detail, as well as the important physical mechanisms that explain how mean-state biases impact the current simulation of nonlinear zonal advection and subsequently the IOD skewness. Section 5 focuses on the ability of models to reproduce the nonlinear vertical advection. Finally, a summary and discussion are provided in section 6.
2. Data and methods
a. Observation/reanalysis and CMIP data
Monthly mean ocean reanalysis data including potential temperature, salinity, and ocean current are taken from the Global Ocean Data Assimilation System (GODAS) of the National Centers for Environmental Prediction (NCEP) (Behringer et al. 1998). Surface heat flux and upper-air wind are taken from ECMWF Reanalysis v5 (ERA5) (Hoffmann et al. 2019). Precipitation is obtained from the NASA/GSFC Global Precipitation Climatology Project (GPCP) version 2.3 (Adler et al. 2003, 2018). The study period is from 1980 to 2020, with anomalies computed as deviations from the detrended climatology for the same period.
The monthly mean outputs of the historical runs from 24 CMIP5 (Taylor et al. 2012) and 20 CMIP6 (Eyring et al. 2016) are used in this study. Since some models contain multiple realizations in their historical runs, we only analyzed the first realization member (r1i1p1 for CMIP5 and r1i1p1f1 for CMIP6) run of each model. The SST, wind, surface heat fluxes, precipitation, potential temperature, salinity, and three-dimensional ocean currents for the periods of 1950–2005 and 1950–2014 are extracted from the historical runs of the CMIP5 and CMIP6 models, respectively. The vertical velocity is not provided in the CMIP5 models and the BCC-CSM2-MR model in CMIP6, and it thus is computed by dividing the vertical mass transport (wmo) by the horizontal area of a grid cell and reference density (1035 kg m−3). In addition, the vertical velocity in the CIESM, CMCC-CM2-SR5, KIOST-ESM, and TaiESM1 models in CMIP6 is calculated using the continuity equation. All the ocean and atmosphere outputs are first interpolated onto a uniform spatial grid of 1° × 1° horizontal resolution before analysis. To remove long-term climate variability, all the data are bandpass filtered to retain variations with periods less than 9 years.
b. IOD identification and model selection
The Dipole Mode Index (DMI) is commonly utilized to identify IOD events in observations (Saji et al. 1999). Upon inspecting CMIP model performance, it is found that the anomaly centers differ from one model to another, making it not possible to use the raw DMI definition for all models. Here a season-reliant empirical orthogonal function (S-EOF) analysis (Wang and An 2005) is applied to identify the major mode of year-to-year SST variability in the Indian Ocean (10°S–5°N, 45°–110°E) from June to November (JJASON). The leading mode, namely S-EOF1 of SSTA during JJASON, performs well in extracting IOD signals with a high contribution rate in both observations and CMIP models (Fig. S1 in the online supplemental material). A positive (negative) IOD event is identified wherever the first principal component (PC1) is greater (less) than its +0.8 (−0.8) standard deviation (STD). Based on the S-EOF1 spatial correlation between observations and models, 24 CMIP5 and 20 CMIP6 models are selected for this study due to their well-simulated IOD patterns (Fig. S2).
c. Skewness
d. Mixed-layer heat budget
e. Convective response
Following Ham (2017), the precipitation response to SST changes is used as a representation of the convective response. Here the climatological precipitation is first binned based on the background SST within 15°S–10°N, 45°–110°E. Then, the anomalous precipitation response by changing SST is calculated with respect to background SST. For a positive SSTA, the convective response is calculated by
3. IOD skewness
a. Simulated IOD asymmetry
Figure 1 depicts the IOD skewness and amplitude in CMIP5/6 models. Notably, six (CAS-ESM2-0, CMCC-CM2-SR5, CMCC-ESM2, E3SM-1-1, EC-Earth3, KIOST-ESM) out of 20 selected CMIP6 models capture a comparable IOD skewness to observations using PC1 (Fig. 1a). It is worth mentioning that these six models were not covered in a prior study (McKenna et al. 2020). However, all CMIP5 models simulate a smaller skewness than observations with the exception of the MPI-ESM-LR model (Fig. 1b), indicating that CMIP6 outperforms CMIP5 in terms of IOD skewness. For the STD of SSTA, most selected models exhibit slightly higher SST variations in the ETIO in contrast to observations, and the CMIP6 multimodel ensemble (MME) mean of SST STD is higher than the counterpart in the CMIP5 ensemble (Fig. 1). Instead of skewness (i.e., the discrepancy between pIOD and nIOD), both enhanced pIOD and nIOD events would contribute to a positive bias in the STD of SSTA. However, due to the diversity of IOD characteristics in CMIP5/6 models, IOD amplitude and skewness show only a weak intermodel correlation. To explore the origins of IOD skewness biases, 20 selected CMIP6 models are categorized into two groups: the more-skewed CMIP6 model group (CAS-ESM2-0, CMCC-CM2-SR5, CMCC-ESM2, E3SM-1-1, EC-Earth3, KIOST-ESM) and the less-skewed CMIP6 model group (remaining selected CMIP6 models). CMIP5 models are also used for comparison.
Skewness of PC1 for representing IOD skewness (orange bars) and SSTA STD over the ETIO (8°S–2°N, 85°–110°E) during JJASON (blue bars) in (a) CMIP6 and (b) CMIP5 models. Orange (blue) line denotes the observed skewness (STD) value. Models with PC1 skewness larger than the observed are marked by gray circles above the skewness bars.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
In the observed surface horizontal skewness pattern during June–November, there is a pronounced negative SST skewness over the ETIO and the area associated with the thermocline dome (Fig. 2a). Regarding the ocean vertical pattern, the negative ocean temperature skewness is mainly observed above 100 m in the ETIO (Figs. 2b,c). Hence, the IOD amplitude asymmetry is predominantly attributed to the ETIO asymmetry, which is seasonally dependent and most noticeable during the mature phase of the IOD. Moreover, the SST skewness center in surface horizontal patterns generally corresponds to the SST STD center, whereas in ocean vertical patterns, the most negative ocean temperature skewness signals are always positioned above the strongest signals of the ocean temperature STD. This is because even a small fluctuation of thermocline would lead to significant temperature changes and produce the strongest STD near the thermocline. However, the IOD skewness is a result of air–sea coupling processes, exhibiting its highest values within the MLD, which is typically shallower than the thermocline. Both CMIP5 and CMIP6 models can reproduce similar SST skewness patterns, characterized by negative signals in the ETIO, but all of the models perform poorly in terms of their ocean vertical skewness (Figs. 2d–l). The more-skewed CMIP6 model group demonstrates the closest resemblance to the observed skewness (Figs. 2g–i), while the other two groups show much smaller skewness values than those in observations and the more-skewed CMIP6 model group. In addition, the negative temperature skewness signals simulated by CMIP models are shallower than those in observations, which might be associated with the simulated shallower thermocline (Ng and Cai 2016).
JJASON skewness (shading) and STD (contours) of (a),(d),(g),(j) SSTA, and anomalous ocean temperature averaged over (b),(e),(h),(k) 5°–5°N and (c),(f),(i),(l) 15°–5°S, based on (a)–(c) observations, (d)–(f) the less-skewed CMIP6 model group, (g)–(i) the more-skewed CMIP6 model group, and (j)–(l) CMIP5 models. The observed and model-simulated 24°, 22° and 20°C isotherms (dash–dotted contours) are also shown in the ocean vertical cross sections. White stippling indicates the 95% significance level based on Student’s t test.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
b. Simulated dynamics of IOD asymmetry
Previous studies revealed that observed IOD asymmetry is mainly caused by the nonlinear zonal (
Composite mixed-layer heat budget during positive and negative IODs during JASO within the ETIO (8°S–2°N, 85°–110°E) in (a) observations, (b) the less-skewed CMIP6 model group, (c) CMIP5 models, and (d) the more-skewed CMIP6 model group. The error bar is based on one STD. A circle over the bar indicates signals passing the 95% significance level.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
The results of the long-term mean NDH within the mixed layer (Fig. 4) are consistent with the composite budget results shown in Fig. 3. The observed NDH exhibits noticeably negative signals in the ETIO (Fig. 4a). The negative long-term mean nonlinear zonal advection is found in the ETIO and southern WTIO (Fig. 4e), and the negative long-term mean nonlinear vertical advection is also observed near the equator and along the Sumatra–Java coast (Fig. 4i). In contrast, the long-term mean nonlinear meridional advection is slightly positive in the southern TIO (Fig. 4m), contributing less to NDH. Both CMIP5 and CMIP6 models are capable of simulating negative NDH in the ETIO, with the largest NDH occurring in the more-skewed CMIP6 model group, followed by the less-skewed CMIP6 model group, and the smallest NDH occurring in the CMIP5 models (Figs. 4b–d). The more-skewed CMIP6 model group simulates comparable nonlinear zonal and vertical advection to observations, while the other two groups reproduce much smaller values, especially the CMIP5 models, indicating improved NDH simulation from CMIP5 to CMIP6 (Figs. 4e–l). These results suggest that both the mean and variability of the mixed-layer NDH are too weak in the less-skewed CMIP6 model group and CMIP5 models, which contributes to a simulated small skewness of the IOD.
Long-term mean (a)–(d) NDH (units: °C month−1), (e)–(h) nonlinear zonal advection, (i)–(l) nonlinear vertical advection, and (m)–(p) nonlinear meridional advection within the mixed layer during JASO based on the (a),(e),(i),(m) observations, (b),(f),(j),(n) less-skewed CMIP6 model group, (c),(g),(k),(o) more-skewed CMIP6 model group, and (d),(h),(l),(p) CMIP5 models. White stippling indicates the 95% significance level based on Student’s t test.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
4. Model biases in nonlinear zonal advection
Figures 5 and 6 show the composite JASO nonlinear zonal advection during pIOD and nIOD, respectively. Consider an initial occurrence of negative SSTA in the ETIO at the beginning of a pIOD event. In response to this anomaly, the generated easterly anomalies trigger Ekman transport that moves the anomalously cold temperature from the east to the west (Figs. 5e,i), resulting in anomalous zonal cold advection and favoring cooling in the ETIO (Fig. 5a). On the contrary, nIOD induces anomalous eastward current and positive anomalous zonal temperature gradient (Figs. 6e,i), as well as causing anomalous zonal cold advection and inhibiting warming in the ETIO (Fig. 6a).
Positive IOD composites during JASO for (a)–(d) nonlinear zonal advection (units: °C month−1), (e)–(h) zonal mixed-layer current anomalies (units: cm s−1), (i)–(l) zonal temperature gradient anomalies within the mixed layer (units: 10−6 °C m−1), and (m)–(p) surface wind anomalies (see scale vector in lower right; units: m s−1; shading is surface zonal wind anomaly), based on the (a),(e),(i),(m) observations, (b),(f),(j),(n) less-skewed CMIP6 model group, (c),(g),(k),(o) more-skewed CMIP6 model group, and (d),(h),(l),(p) CMIP5 models. White stippling indicates the 95% significance level based on Student’s t test. Wind vectors in (m)–(p) are shown when exceeding the 95% significance level based on Student’s t test.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
As in Fig. 5, but for negative IOD.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
The nonlinear zonal advection patterns during pIOD are similar across all CMIP groups, but only the more-skewed CMIP6 group matches the observed variability (Figs. 5a–d). This depends on how well the models simulate the zonal current and temperature gradient anomalies in the mixed layer, especially the zonal current anomalies (Figs. 5f–h,j–l). The more-skewed CMIP6 model group produces larger anomalies in zonal current and zonal temperature gradient due to stronger easterly wind anomalies (Fig. 5o). The less-skewed CMIP6 group and the CMIP5 models, on the other hand, have weaker atmospheric responses than the more-skewed CMIP6 model group and observations (Figs. 5m–p). For nIOD, the nonlinear zonal advection and the westerly wind anomalies in CMIP6 are consistent with observations (Figs. 6a–c,m–o). The CMIP5 models show the smallest nonlinear zonal advection, which is related to their weak surface wind anomalies (Figs. 6d,p). In conclusion, nonlinear zonal advection biases are larger during pIOD than nIOD and the simulation of nonlinear zonal advection is improved from CMIP5 to CMIP6.
The simulations of atmospheric response to SST changes during IOD events are reflected in the simulated surface zonal wind anomalies. The asymmetric atmospheric response has been attributed to the nonlinear relationship between SST and precipitation by previous studies (Hoerling et al. 1997; Ohba and Ueda 2009). In observations, increasing SST typically causes an increase in precipitation if SSTs are lower than around 29°C (Fig. 7a), implying that the convective response to SST forcing is strongly dependent on the background SST. The observed SST for maximum precipitation sensitivity to positive SSTA is around 27°C and to negative SSTA is around 27.5°C (Fig. 7e). Most of the CMIP models can reproduce similar precipitation response to SSTA as observations, with positive (negative) net convective response lower (greater) than 27.5°C (Figs. 7e–h).
(a)–(d) Average precipitation (units: mm day−1) with respect to SST (units: °C) and (e)–(h) precipitation responses to SST changes (units: mm day−1 0.5°C−1), based on (a),(e) observations, (b),(f) less-skewed CMIP6 model group, (c),(g) more-skewed CMIP6 model group, and (d),(h) CMIP5 models. Red (blue) line gives precipitation change due to a 0.5°C increase (decrease) in SST and black bar is net precipitation response. (i) Mean-state SST (shading; units: °C), precipitation (contour; in intervals of 3 mm day−1), and 850-hPa wind (see scale vector in lower right; units: m s−1) and (m) net precipitation response (units: mm day−1 0.5°C−1) during JASO in observations. (j)–(l) Mean-state biases of SST, precipitation and 850-hPa wind and (n)–(p) net precipitation response biases, based on the (j),(n) less-skewed CMIP6 model group, (k),(o) more-skewed CMIP6 model group and (l),(p) CMIP5 models. White stippling indicates the 95% significance level based on Student’s t test. (q) Schematic diagram illustrating how mean-state positive IOD-like biases affect atmospheric response to SSTA. In observation, the WTIO (ETIO) captures larger precipitation response to positive (negative) SSTA than to negative (positive) SSTA, favoring the pIOD with stronger atmospheric response. In models with mean-state positive IOD-like biases, a warmer WTIO results in less (more) precipitation response to positive (negative) SSTA and a colder ETIO causes less negative net precipitation response, leading to weaker atmospheric response during pIOD, less favor of pIOD and less IOD asymmetry.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
To understand how asymmetric precipitation changes with SSTA, we can compute the net convective response from a given mean-state SST. Figure 7i shows the observed mean-state SST during JASO, and Fig. 7m shows its corresponding net convective response. The observations reveal that the WTIO has a mean-state SST of about 27°C during JASO, which means that positive SSTAs trigger stronger convection than negative SSTAs, resulting in a positive net convective response. On the other hand, the ETIO has a mean-state SST of about 28.5°C, which means that negative SSTAs induce more precipitation anomalies than positive SSTAs, resulting in a negative net convective response. In the models with significantly mean-state positive IOD-like biases, the WTIO becomes warmer and less sensitive to positive SSTAs, leading to less positive net convective response. While the ETIO becomes colder and induces less negative net convective response. For more-skewed CMIP6 models, there is weaker warm WTIO bias and minor ETIO bias, accompanied by less convective response biases except the slightly negative net convective response bias in the WTIO. Thus, these models can simulate comparable atmospheric response and nonlinear zonal advection during pIOD. In contrast, mean-state positive IOD-like biases are evident in both the less-skewed CMIP6 model group and CMIP5 models (Figs. 7j,l), resulting in less positive (negative) net convective response in the WTIO (SETIO) (Figs. 7n,p). These results imply that WTIO (SETIO) in these less-skewed models is less sensitive to positive (negative) SSTA than observations, leading to less atmospheric response to SSTA and weaker efficiency in the Bjerknes feedback during pIOD. In conclusion, mean-state positive IOD-like biases contribute to less convective response and less negative nonlinear zonal advection during pIOD.
Intermodel statistics of CMIP6 also support that the simulated less IOD skewness is associated with the mean-state positive IOD-like biases. Among the CMIP6 models, models with larger mean-state SST gradient during JASO in the TIO tend to have a less positive IOD skewness, with a correlation coefficient of −0.70 (passes the 95% significance level) (Fig. 8a). Similarly, the models with a weaker mean-state westerly wind in the TIO during JASO, which also implies stronger positive IOD-like biases, tend to have a less positive IOD skewness, with a correlation coefficient of 0.67 (significant at the 95% level) (Fig. 8b). As mentioned above, mean-state biases would affect nonlinear zonal advection by modulating the sensitivity of atmospheric response to SST changes. Weaker mean-state westerlies in the TIO during JASO (i.e., stronger mean-state positive IOD-like biases) reduce the mean-state negative nonlinear zonal advection by decreasing the west–east net convective response gradient. Notably, the west–east net convective response gradient is highly correlated with the IOD skewness, with a correlation coefficient of 0.68 (passes the 95% significance level) (Fig. 8c).
Mean state in observations and CMIP6 models. (a) SST gradient (units: °C; 45°–60°E minus 80°–100°E, 5°S–5°N; JASO) vs IOD skewness. (b) Surface zonal wind (units: m s−1; 5°S–5°N, 60°–90°E; JASO) vs IOD skewness. (c) Net convective response gradient (units: mm day−1 0.5°C−1; 45°–60°E minus 80°–100°E, 5°S–5°N; JASO) vs IOD skewness. (d) Surface meridional wind (units: m s−1; 5°S–5°N, 45°–60°E; JJA) vs SST gradient (units: °C; 45°–60°E minus 80°–100°E, 5°S–5°N; JJA). (e) SST gradient (units: °C; 45°–60°E minus 80°–100°E, 5°S–5°N; JJA) vs surface zonal wind (units: m s−1; 5°S–5°N, 60°–90°E; SON). (f) Surface meridional wind (units: m s−1; 5°S–5°N, 45°–60°E; JJA) vs surface zonal wind (units: m s−1; 5°S–5°N, 60°–90°E; SON). (g) Surface meridional wind (units: m s−1; 10°S–0°, 90°–100°E; JJA) vs SST gradient (units: °C; 45°–60°E minus 80°–100°E, 5°S–5°N; JJA). (h) Surface meridional wind (units: m s−1; 10°S–0°, 90°–100°E; JJA) vs surface zonal wind (units: m s−1; 5°S–5°N, 60°–90°E; SON). Blue line is the least squares linear fit. The r value is the correlation coefficient.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
Previous studies have mentioned that CMIP models frequently exhibit positive IOD-like biases in the mean state, which is associated with the weaker simulation of Indian summer monsoon (Li et al. 2015; Long et al. 2020). Our results also support the hypothesis that the surface meridional wind bias in the WTIO is responsible for the seasonally dependent IOD-like biases. Models with a weaker summer monsoon in June–August (JJA) tend to develop a larger TIO SST gradient bias in JJA, with a correlation coefficient of −0.60 (Fig. 8d). The mean-state TIO SST gradient in JJA and the mean-state TIO surface zonal wind in September–November (SON) are correlated at 0.78, suggesting that JJA SST gradient bias drives the SON easterly bias in the TIO (Fig. 8e). Weaker simulated Indian summer monsoon leads to warmer mean condition of WTIO in JJA; at the same time, Bjerknes feedback amplifies the TIO SST gradient and easterly biases in SON (Fig. 8f). On the other hand, the simulation of surface meridional wind near the Sumatra–Java coast likewise promotes the growth of positive IOD-like biases. The presence of stronger mean-state southeasterly wind in JJA leads to the cooling of ETIO through the wind-forced Ekman transport, thus facilitating the formation of a more pronounced easterly bias in SON (Figs. 8g,h).
Figure 9 depicts the MME patterns of SST, precipitation, and 850-hPa wind biases. These patterns provide additional insight into the development of IOD-like biases from June to October. The less-skewed CMIP6 model group and CMIP5 models show stronger positive IOD-like biases than the more-skewed CMIP6 model group, with much warmer WTIO, cooler ETIO, and stronger TIO easterly bias (Figs. 9f–t). In June, weaker WTIO southerly wind generates warmer WTIO in the less-skewed models (Figs. 9f,p). Through the Bjerknes feedback, this WTIO warm bias grows and peaks during July–September, leading to significant ETIO cold bias and stronger TIO easterly bias during August–October (Figs. 9g–j,q–t). In addition, the southeasterly wind bias near the Sumatra–Java coast in summer also helps the development of the ETIO cold bias and the TIO easterly bias (Figs. 9f,g,p,q). In summary, weaker Indian summer monsoon and stronger Sumatra–Java coastal wind cause mean-state positive IOD-like biases that reduce the positive (negative) net convective response in the WTIO (ETIO), leading to weaker easterly anomalies during pIOD, less negative mixed-layer nonlinear zonal advection, and ultimately less simulated IOD skewness in models.
(a)–(e) Mean-state SST (shading; units: °C), precipitation (contour; in intervals of 3 mm day−1), and 850-hPa wind (see scale vector in lower right; units: m s−1) from June to October in observations. Mean-state biases of SST, precipitation, and 850-hPa wind based on the (f)–(j) less-skewed CMIP6 model group, (k)–(o) more-skewed CMIP6 model group, and (p)–(t) CMIP5 models. Biases of SST (white stippling), precipitation and wind are shown when exceeding the 95% significance level based on Student’s t test.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
5. Model biases in nonlinear vertical advection
Vertical nonlinear advection is another important factor in the IOD. Figures 10 and 11 show the composite patterns of nonlinear vertical advection, vertical velocity anomalies, and vertical temperature gradient anomalies at the base of mixed layer during pIOD and nIOD, respectively. In observations, both anomalous easterly and southerly winds play a role in contributing to anomalous upwelling at the equator and near the Sumatra–Java coast during pIOD (Fig. 10e), leading to positive anomalous vertical temperature gradient and cold nonlinear vertical advection (Figs. 10a,i). Note that the maximum temperature amplitude is located at the subsurface layer so that the MLT anomalies are smaller than those below the mixed layer, forming a positive anomalous vertical temperature gradient in pIOD. Negative nonlinear vertical advection emerges in nIOD as well (Fig. 11a). The response of vertical motion to wind anomaly tends to be linear (Figs. 10e and 11e), while the response of vertical temperature gradient is nonlinear (Figs. 10i and 11i) and it enlarges the negative nonlinear vertical advection during pIOD, favoring the IOD asymmetry.
Positive IOD composites during JASO for (a)–(d) nonlinear vertical advection (units: °C month−1), (e)–(h) vertical velocity anomalies (units: 10−3 cm s−1), and (i)–(l) vertical temperature gradient anomalies at the base of mixed layer (units: °C m−1), based on (a),(e),(i) observations, (b),(f),(j) less-skewed CMIP6 model group, (c),(g),(k) more-skewed CMIP6 model group, and (d),(h),(l) CMIP5 models. White stippling indicates the 95% significance level based on Student’s t test.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
As in Fig. 10, but for negative IOD.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
Observations and CMIP5/6 models share similar composite patterns of nonlinear vertical advection and its associated variables. During pIOD, the ETIO nonlinear vertical advection in the more-skewed CMIP6 model group is comparable to observations and notably larger than that in the less-skewed CMIP6 model group (Figs. 10a–c). This is mainly due to the larger simulations of anomalous vertical temperature gradient (Figs. 10i–k). In terms of nIOD, both CMIP6 groups can capture variations in nonlinear vertical advection as in observations, with slightly higher values in the more-skewed CMIP6 model group (Figs. 11a–c). The CMIP5 models have the weakest MME nonlinear vertical advection (Figs. 10d and 11d), which is related to the small variations in vertical motion (Figs. 10h and 11h). Since there is no vertical velocity (wo) variable output in CMIP5 models, vertical motion is derived from vertical mass transport (wmo), which may introduce biases in the derived vertical motion.
The nonlinear response of SST to thermocline changes is important in developing positive IOD skewness in observations. Sea level anomaly (SLA) versus SSTA in the ETIO (Fig. 12a) illustrates this nonlinear relationship (α = −0.15; α is the nonlinear coefficient of a quadratic function), indicating that ETIO SST is more sensitive to a shallower thermocline. In comparison to observations, most of the CMIP5/6 models exhibit reduced nonlinear SST response to the thermocline changes. Only the more-skewed CMIP6 model group can capture the nonlinear relationship between SLA and SSTA (α = −0.13) (Fig. 12c), indicating improved simulation of this nonlinear relationship in these CMIP6 models. In the other two groups, the relationship between SLA and SSTA remains linear (Figs. 12b,d), implying a smaller SST response to a shallower thermocline.
Standardized monthly SLA (8°S–2°N, 85°–110°E) vs SSTA (8°S–2°N, 85°–110°E) based on (a) observations, (b) less-skewed CMIP6 model group, (c) more-skewed CMIP6 model group, and (d) CMIP5 models. Red (blue) crosses refer to positive (negative) IOD during JJASON; (b)–(d) consist of data during IOD events only. Black curve gives a quadratic fit of SLA and SSTA, and α is the nonlinear coefficient.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
The nonlinear SST response to thermocline changes is tightly related to the background upper-ocean temperature. Previous research using CMIP5 models showed that IOD skewness is caused by asymmetric thermocline feedback, which is associated with the mean-state simulation of thermocline (Ng et al. 2014; Ng and Cai 2016). Here, we find that ocean stratification in the ETIO accounts for simulated IOD skewness by affecting the MLT response to thermocline changes. In fact, ocean stratification in the ETIO is positively correlated with simulated IOD skewness (r = 0.70) (Fig. 13a), whereas the thermocline (defined by the depth of 20°C isotherm, Z20) is negatively correlated with the simulated IOD skewness (r = −0.53) (Fig. 13b) among selected CMIP6 models. An enhanced ocean stratification contributes to a stronger anomalous vertical temperature gradient. Although a shallower thermocline may sometimes weaken the asymmetry of thermocline–SST feedback (Cai et al. 2013), a stronger MLT response to thermocline changes would enlarge the anomalous vertical temperature gradient during both pIOD and nIOD, all leading to more negative nonlinear vertical advection and positive IOD skewness.
Mean ocean state in observations and CMIP6 models. (a) Ocean stratification (units: °C; 8°S–2°N, 85°–110°E; JASO) vs IOD skewness. (b) Thermocline (Z20; units: m; 8°S–2°N, 85°–110°E; JASO) vs IOD skewness. Blue line is the least squares linear fit. The r value is the correlation coefficient. The ocean stratification is calculated as the difference between the mean temperature over the upper 60 m and the temperature averaged from 90 to 120 m.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
Figure 14 depicts the vertical patterns of MLD and temperature from June to October. The less-skewed CMIP6 model group and CMIP5 models have similar vertical temperature patterns, with a deeper MLD in the TIO, deeper thermocline in the WTIO, significant warm bias in the WTIO, and slightly cold bias in the ETIO between 60- and 150-m depth (Figs. 14f–j,p–t). These biases are consistent with the strong positive IOD-like biases in the mean state. Through the Bjerknes feedback loop, an easterly wind bias develops along the equator, causing unrealistic thermocline tilting and significantly warmer (colder) mean temperatures over the upper WTIO (ETIO) in SON. On the other hand, the more-skewed CMIP6 model group has more realistic thermocline depths in most regions, except for a slightly shallower thermocline in the ETIO (Figs. 14k–o). Despite having larger cold temperature biases within the ETIO, the more-skewed CMIP6 model group can simulate more realistic nonlinear vertical advection with the help of stronger ocean stratification. Increased mean-state ocean stratification promotes larger anomalous vertical temperature gradients during IOD, resulting in more negative nonlinear vertical advection and larger IOD skewness. Therefore, better simulations of upper-ocean temperature and oceanic responses in the ETIO are important to reproduce a realistic IOD skewness.
(a)–(e) Mean-state vertical cross sections of temperature (units: °C) averaged over 8°S–2°N from June to October in observations. Temperature biases based on the (f)–(j) less-skewed CMIP6 model group, (k)–(o) more-skewed CMIP6 model group, and (p)–(t) CMIP5 models. Asterisk (circle) denotes observed (simulated) MLD. Black dash-dotted (solid) contour is observed (simulated) Z20. White stippling indicates the 95% significance level based on Student’s t test.
Citation: Journal of Climate 37, 3; 10.1175/JCLI-D-23-0412.1
Nonlinear meridional advection acts as the damping term in the negative NDH. Figures S3 and S4 show composite patterns of nonlinear meridional advection during pIOD and nIOD, respectively. The observed nonlinear meridional advection during IOD is small (less than 0.2°C month−1), with positive values in the SETIO and somewhat negative values near the equator (Figs. S3a and S4a). In CMIP5/6 models, nonlinear meridional advection is also a minor contributor to NDH (Figs. S3b–d and S4b–d). During pIOD (nIOD), CMIP models simulate significant positive (negative) anomalous meridional temperature gradient in the northern ETIO (NETIO) and negative (positive) gradient in the SETIO (Figs. S3j–l and S4j–l). These meridional temperature gradient anomalies are comparable to or slightly larger than the anomalous zonal temperature gradient. However, since the anomalous zonal current is greatly boosted by the Bjerknes feedback, it is nearly 4 times larger than the anomalous meridional current. Hence, nonlinear meridional advection can be easily offset by the other two NDH terms. Additionally, even though the nonlinear meridional advection during pIOD in the more-skewed CMIP6 model group is greater than observations (Figs. S3a,c), the negative anomaly in the NETIO and positive anomaly in the SETIO tend to neutralize each other, reducing their net impact on NDH.
6. Summary and discussion
The IOD is a climate phenomenon that affects weather patterns around the Indian Ocean–rim countries and even worldwide. The IOD exhibits a noticeable amplitude asymmetry, which is particularly evident during extreme IOD events that occur more frequently during positive phases, triggering severe droughts, floods, and wildfires in different regions. In this study, we evaluate the present-day simulation of IOD skewness in CMIP5/6 models and find that most of the recent coupled general circulation models tend to simulate a weaker IOD skewness. Compared to observations, nearly all the selected CMIP5 models produce a smaller IOD skewness except the MPI-ESM-LR model, but 6 out of 20 selected CMIP6 models can simulate a realistic IOD skewness, indicating some improvement in IOD skewness from CMIP5 to CMIP6. To explore the origins of underestimated IOD skewness, the selected CMIP6 models are classified into two groups: the less-skewed CMIP6 model group and the more-skewed CMIP6 model group. Models in the more-skewed CMIP6 model group simulate comparable IOD skewness with observations.
The amplitude asymmetry of observed IOD events is in relation to negative NDH, mainly owing to nonlinear zonal and vertical advection (Hong et al. 2008a,b). In CMIP5/6 models, we find that the NDH simulation is essential to the simulation of IOD skewness. Only the more-skewed CMIP6 model group simulates NDH that is comparable to observations. In the other two less-skewed groups (i.e., the CMIP5 model group and less-skewed CMIP6 model group), the simulated NDH is much smaller, which is mainly due to the simulated biases in nonlinear zonal and vertical advection, but the biases in nonlinear meridional advection contribute less to simulated NDH.
Simulated biases of nonlinear zonal advection are related to the mean-state positive IOD-like biases. In observations, the background WTIO SST during JASO is around 27°C, with positive net convective response, while the ETIO SST is around 28.5°C, with negative net convective response. Models with warmer mean-state WTIO weaken the WTIO’s positive net convective response to SSTAs, reducing atmospheric response during pIOD. This leads to weaker surface easterly anomalies during pIOD, which in turn results in weaker ETIO nonlinear zonal advection and less positive IOD skewness. Meanwhile, colder mean-state ETIO induces less negative net convective response, which is also less favorable for the development of pIOD. The more-skewed CMIP6 model group shows fewer IOD-like biases than the other two groups and thereby simulates more realistic IOD skewness. Furthermore, the positive IOD-like biases are caused by the weaker cross-equatorial summer monsoon in the WTIO and the summer southeasterly wind bias near the Sumatra–Java coast.
Nonlinear vertical advection is another key factor for the IOD asymmetry. Only the more-skewed CMIP6 model group achieves comparable values to observations, which is related to the ETIO stratification. Stronger ocean stratification causes stronger MLT response to thermocline changes, which increases the anomalous vertical temperature gradient and more negative nonlinear vertical advection, favoring the IOD asymmetry. Despite having larger cold biases in the ETIO, the more-skewed CMIP6 model group reproduces better temperature responses and nonlinear vertical advection. These findings imply that models with better simulations of Indian summer monsoon and ETIO oceanic responses perform well in replicating present-day IOD skewness. Therefore, substantial efforts should be undertaken to reduce the monsoon bias as well as to advance the upper-ocean temperature simulation.
This study mainly focuses on the effect of NDH biases on the simulated IOD skewness. Some thermodynamic processes would also partially contribute to the IOD skewness, such as the SCR feedback (Hong et al. 2008a) and the latent heat flux response in the ETIO (Fig. 3). In observations, notwithstanding that there is no clear evidence linking IOD skewness to the asymmetry of remote ENSO forcing in summer (Hong et al. 2008a), Du et al. (2020) found that the Rossby wave caused by El Niño in the spring may enable Bjerknes feedback to occur earlier and facilitate the emergence of the 2019 extreme IOD event. The Indonesian Throughflow, which is closely related to the ENSO (Meyers 1996), could impact the barrier layer in the ETIO and potentially enhance the IOD asymmetry (Cai et al. 2013). Recently An et al. (2023) suggested that ENSO may play a secondary role in IOD asymmetry, alongside local nonlinear processes, as demonstrated by a nonlinear IOD model. Like the IOD, CMIP models also underestimate ENSO skewness (McKenna et al. 2020; Hayashi et al. 2020). The question arises as to whether these biases in ENSO skewness within the models have implications for the accurate simulation of IOD skewness. Therefore, further research is needed to investigate how simulations of pan-tropical interactions (Cai et al. 2019) contribute to the simulated IOD skewness. On the other hand, it is important to take into account how biases in IOD skewness simulations may affect the simulations of the present-day IOD climate impact. Meanwhile, some studies projected that extreme IOD events would occur more frequently as a result of global warming (Cai et al. 2014, 2021); however, it is not yet known how the present-day simulation biases of IOD skewness will affect future IOD changes (Ng and Cai 2016; Wang et al. 2021). Our findings help to improve our understanding of the present-day model’s performances and constitute a step in assessing their predictions for the upcoming changes.
Acknowledgments.
Funding was received from the National Natural Science Foundation of China (42076020), Guangdong Basic and Applied Basic Research Foundation (2023B1515020009), and Youth Innovation Promotion Association CAS (2020340).
Data availability statement.
The GODAS data can be obtained from https://psl.noaa.gov/data/gridded/data.godas.html. The ERA5 data are taken from https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels-monthly-means?tab=overview. The GPCP data are derived from https://www.ncei.noaa.gov/access/metadata/landing-page/bin/iso?id=gov.noaa.ncdc:C00979. The CMIP5 data can be obtained from https://esgf-node.llnl.gov/search/cmip5/. The CMIP6 data are taken from https://esgf-node.llnl.gov/search/cmip6/.
REFERENCES
Abram, N. J., and Coauthors, 2020: Coupling of Indo-Pacific climate variability over the last millennium. Nature, 579, 385–392, https://doi.org/10.1038/s41586-020-2084-4.
Adler, R. F., and Coauthors, 2003: The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 1147–1167, https://doi.org/10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.
Adler, R. F., and Coauthors, 2018: The Global Precipitation Climatology Project (GPCP) monthly analysis (new version 2.3) and a review of 2017 global precipitation. Atmosphere, 9, 138, https://doi.org/10.3390/atmos9040138.
An, S.-I., H.-J. Park, S.-K. Kim, W. Cai, A. Santoso, D. Kim, and J.-S. Kug, 2023: Main drivers of Indian Ocean dipole asymmetry revealed by a simple IOD model. npj Climate Atmos. Sci., 6, 93, https://doi.org/10.1038/s41612-023-00422-2.
Annamalai, H., R. Murtugudde, J. Potemra, S. P. Xie, P. Liu, and B. Wang, 2003: Coupled dynamics over the Indian Ocean: Spring initiation of the zonal mode. Deep-Sea Res. II, 50, 2305–2330, https://doi.org/10.1016/S0967-0645(03)00058-4.
Behringer, D. W., M. Ji, and A. Leetmaa, 1998: An improved coupled model for ENSO prediction and implications for ocean initialization. Part I: The ocean data assimilation system. Mon. Wea. Rev., 126, 1013–1021, https://doi.org/10.1175/1520-0493(1998)126<1013:AICMFE>2.0.CO;2.
Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97, 163–172, https://doi.org/10.1175/1520-0493(1969)097<0163:ATFTEP>2.3.CO;2.
Cai, W., and Y. Qiu, 2013: An observation-based assessment of nonlinear feedback processes associated with the Indian Ocean dipole. J. Climate, 26, 2880–2890, https://doi.org/10.1175/JCLI-D-12-00483.1.
Cai, W., T. Cowan, and M. Raupach, 2009: Positive Indian Ocean dipole events precondition southeast Australia bushfires. Geophys. Res. Lett., 36, L19710, https://doi.org/10.1029/2009GL039902.
Cai, W., P. van Rensch, T. Cowan, and H. H. Hendon, 2012: An asymmetry in the IOD and ENSO teleconnection pathway and its impact on Australian climate. J. Climate, 25, 6318–6329, https://doi.org/10.1175/JCLI-D-11-00501.1.
Cai, W., X.-T. Zheng, E. Weller, M. Collins, T. Cowan, M. Lengaigne, W. Yu, and T. Yamagata, 2013: Projected response of the Indian Ocean dipole to greenhouse warming. Nat. Geosci., 6, 999–1007, https://doi.org/10.1038/ngeo2009.
Cai, W., A. Santoso, G. Wang, E. Weller, L. Wu, K. Ashok, Y. Masumoto, and T. Yamagata, 2014: Increased frequency of extreme Indian Ocean dipole events due to greenhouse warming. Nature, 510, 254–258, https://doi.org/10.1038/nature13327.
Cai, W., and Coauthors, 2019: Pantropical climate interactions. Science, 363, eaav4236, https://doi.org/10.1126/science.aav4236.
Cai, W., K. Yang, L. Wu, G. Huang, A. Santoso, B. Ng, G. Wang, and T. Yamagata, 2021: Opposite response of strong and moderate positive Indian Ocean dipole to global warming. Nat. Climate Change, 11, 27–32, https://doi.org/10.1038/s41558-020-00943-1.
de Boyer Montégut, C., J. Vialard, S. S. C. Shenoi, D. Shankar, F. Durand, C. Ethé, and G. Madec, 2007: Simulated seasonal and interannual variability of the mixed layer heat budget in the northern Indian Ocean. J. Climate, 20, 3249–3268, https://doi.org/10.1175/JCLI4148.1.
Du, Y., W. Cai, and Y. Wu, 2013: A new type of the Indian Ocean dipole since the mid-1970s. J. Climate, 26, 959–972, https://doi.org/10.1175/JCLI-D-12-00047.1.
Du, Y., Y. Zhang, L.-Y. Zhang, T. Tozuka, B. Ng, and W. Cai, 2020: Thermocline warming induced extreme Indian Ocean dipole in 2019. Geophys. Res. Lett., 47, e2020GL090079, https://doi.org/10.1029/2020GL090079.
Endo, S., and T. Tozuka, 2016: Two flavors of the Indian Ocean dipole. Climate Dyn., 46, 3371–3385, https://doi.org/10.1007/s00382-015-2773-0.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Guan, Z., and T. Yamagata, 2003: The unusual summer of 1994 in East Asia: IOD teleconnections. Geophys. Res. Lett., 30, 1544, https://doi.org/10.1029/2002GL016831.
Ham, Y.-G., 2017: A reduction in the asymmetry of ENSO amplitude due to global warming: The role of atmospheric feedback. Geophys. Res. Lett., 44, 8576–8584, https://doi.org/10.1002/2017GL074842.
Hayashi, M., F.-F. Jin, and M. F. Stuecker, 2020: Dynamics for El Niño–La Niña asymmetry constrain equatorial-Pacific warming pattern. Nat. Commun., 11, 4230, https://doi.org/10.1038/s41467-020-17983-y.
Hoerling, M. P., A. Kumar, and M. Zhong, 1997: El Niño, La Niña, and the nonlinearity of their teleconnections. J. Climate, 10, 1769–1786, https://doi.org/10.1175/1520-0442(1997)010<1769:ENOLNA>2.0.CO;2.
Hoffmann, L., and Coauthors, 2019: From ERA-Interim to ERA5: The considerable impact of ECMWF’s next-generation reanalysis on Lagrangian transport simulations. Atmos. Chem. Phys., 19, 3097–3124, https://doi.org/10.5194/acp-19-3097-2019.
Hong, C.-C., and T. Li, 2010: Independence of SST skewness from thermocline feedback in the eastern equatorial Indian Ocean. Geophys. Res. Lett., 37, L11702, https://doi.org/10.1029/2010GL043380.
Hong, C.-C., T. Li, LinHo, and J.-S. Kug, 2008a: Asymmetry of the Indian Ocean dipole. Part I: Observational analysis. J. Climate, 21, 4834–4848, https://doi.org/10.1175/2008JCLI2222.1.
Hong, C.-C., T. Li, and J.-J. Luo, 2008b: Asymmetry of the Indian Ocean dipole. Part II: Model diagnosis. J. Climate, 21, 4849–4858, https://doi.org/10.1175/2008JCLI2223.1.
Jourdain, N. C., A. Sen Gupta, A. S. Taschetto, C. C. Ummenhofer, A. F. Moise, and K. Ashok, 2013: The Indo-Australian monsoon and its relationship to ENSO and IOD in reanalysis data and the CMIP3/CMIP5 simulations. Climate Dyn., 41, 3073–3102, https://doi.org/10.1007/s00382-013-1676-1.
Kara, A. B., P. A. Rochford, and H. E. Hurlburt, 2000: An optimal definition for ocean mixed layer depth. J. Geophys. Res., 105, 16 803–16 821, https://doi.org/10.1029/2000JC900072.
Li, G., S.-P. Xie, and Y. Du, 2015: Monsoon-induced biases of climate models over the tropical Indian Ocean. J. Climate, 28, 3058–3072, https://doi.org/10.1175/JCLI-D-14-00740.1.
Li, T., B. Wang, C.-P. Chang, and Y. Zhang, 2003: A theory for the Indian Ocean dipole–zonal mode. J. Atmos. Sci., 60, 2119–2135, https://doi.org/10.1175/1520-0469(2003)060<2119:ATFTIO>2.0.CO;2.
Long, S.-M., G. Li, K. Hu, and J. Ying, 2020: Origins of the IOD-like biases in CMIP multimodel ensembles: The atmospheric component and ocean–atmosphere coupling. J. Climate, 33, 10 437–10 453, https://doi.org/10.1175/JCLI-D-20-0459.1.
Luo, J.-J., S. Behera, Y. Masumoto, H. Sakuma, and T. Yamagata, 2008: Successful prediction of the consecutive IOD in 2006 and 2007. Geophys. Res. Lett., 35, L14S02, https://doi.org/10.1029/2007GL032793.
Masson, S., J.-P. Boulanger, C. Menkes, P. Delecluse, and T. Yamagata, 2004: Impact of salinity on the 1997 Indian Ocean dipole event in a numerical experiment. J. Geophys. Res., 109, C02002, https://doi.org/10.1029/2003JC001807.
McKenna, S., A. Santoso, A. Sen Gupta, A. S. Taschetto, and W. Cai, 2020: Indian Ocean dipole in CMIP5 and CMIP6: Characteristics, biases, and links to ENSO. Sci. Rep., 10, 11500, https://doi.org/10.1038/s41598-020-68268-9.
Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J. F. B. Mitchell, R. J. Stouffer, and K. E. Taylor, 2007: THE WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 1383–1394, https://doi.org/10.1175/BAMS-88-9-1383.
Meyers, G., 1996: Variation of Indonesian throughflow and the El Niño-Southern Oscillation. J. Geophys. Res., 101, 12 255–12 263, https://doi.org/10.1029/95JC03729.
Nakazato, M., S. Kido, and T. Tozuka, 2021: Mechanisms of asymmetry in sea surface temperature anomalies associated with the Indian Ocean dipole revealed by closed heat budget. Sci. Rep., 11, 22546, https://doi.org/10.1038/s41598-021-01619-2.
Ng, B., and W. Cai, 2016: Present-day zonal wind influences projected Indian Ocean dipole skewness. Geophys. Res. Lett., 43, 11 392–11 399, https://doi.org/10.1002/2016GL071208.
Ng, B., W. Cai, and K. Walsh, 2014: The role of the SST–thermocline relationship in Indian Ocean dipole skewness and its response to global warming. Sci. Rep., 4, 6034, https://doi.org/10.1038/srep06034.
Ogata, T., S.-P. Xie, J. Lan, and X. Zheng, 2013: Importance of ocean dynamics for the skewness of the Indian Ocean dipole mode. J. Climate, 26, 2145–2159, https://doi.org/10.1175/JCLI-D-11-00615.1.
Ohba, M., and H. Ueda, 2009: Role of nonlinear atmospheric response to SST on the asymmetric transition process of ENSO. J. Climate, 22, 177–192, https://doi.org/10.1175/2008JCLI2334.1.
Page, S. E., F. Siegert, J. O. Rieley, H.-D. V. Boehm, A. Jaya, and S. Limin, 2002: The amount of carbon released from peat and forest fires in Indonesia during 1997. Nature, 420, 61–65, https://doi.org/10.1038/nature01131.
Qiu, Y., W. Cai, L. Li, and X. Guo, 2012: Argo profiles variability of barrier layer in the tropical Indian Ocean and its relationship with the Indian Ocean dipole. Geophys. Res. Lett., 39, L08605, https://doi.org/10.1029/2012GL051441.
Qiu, Y., W. Cai, X. Guo, and B. Ng, 2014: The asymmetric influence of the positive and negative IOD events on China’s rainfall. Sci. Rep., 4, 4943, https://doi.org/10.1038/srep04943.
Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360–363, https://doi.org/10.1038/43854.
Schott, F. A., S.-P. Xie, and J. P. McCreary Jr., 2009: Indian Ocean circulation and climate variability. Rev. Geophys., 47, RG1002, https://doi.org/10.1029/2007RG000245.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, https://doi.org/10.1175/BAMS-D-11-00094.1.
Tozuka, T., J.-J. Luo, S. Masson, and T. Yamagata, 2008: Tropical Indian Ocean variability revealed by self-organizing maps. Climate Dyn., 31, 333–343, https://doi.org/10.1007/s00382-007-0356-4.
Vecchi, G. A., and B. J. Soden, 2007: Global warming and the weakening of the tropical circulation. J. Climate, 20, 4316–4340, https://doi.org/10.1175/JCLI4258.1.
Vecchi, G. A., B. J. Soden, A. T. Wittenberg, I. M. Held, A. Leetmaa, and M. J. Harrison, 2006: Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature, 441, 73–76, https://doi.org/10.1038/nature04744.
Wang, B., and S.-I. An, 2005: A method for detecting season-dependent modes of climate variability: S-EOF analysis. Geophys. Res. Lett., 32, L15710, https://doi.org/10.1029/2005GL022709.
Wang, G., W. Cai, and A. Santoso, 2021: Simulated thermocline tilt over the tropical Indian Ocean and its influence on future sea surface temperature variability. Geophys. Res. Lett., 48, e2020GL091902, https://doi.org/10.1029/2020GL091902.
Webster, P. J., A. M. Moore, J. P. Loschnigg, and R. R. Leben, 1999: Coupled ocean–atmosphere dynamics in the Indian Ocean during 1997–98. Nature, 401, 356–360, https://doi.org/10.1038/43848.
Xie, S.-P., and S. G. H. Philander, 1994: A coupled ocean-atmosphere model of relevance to the ITCZ in the eastern Pacific. Tellus, 46A, 340–350, https://doi.org/10.3402/tellusa.v46i4.15484.
Xie, S.-P., Y. Du, G. Huang, X.-T. Zheng, H. Tokinaga, K. Hu, and Q. Liu, 2010: Decadal shift in El Niño influences on Indo–western Pacific and East Asian climate in the 1970s. J. Climate, 23, 3352–3368, https://doi.org/10.1175/2010JCLI3429.1.
Xu, K., H.-Y. Miao, B. Liu, C.-Y. Tam, and W. Wang, 2020: Aggravation of record-breaking drought over the mid-to-lower reaches of the Yangtze River in the post-monsoon season of 2019 by anomalous Indo-Pacific oceanic conditions. Geophys. Res. Lett., 47, e2020GL090847, https://doi.org/10.1029/2020GL090847.
Yang, K., W. Cai, G. Huang, G. Wang, B. Ng, and S. Li, 2020: Oceanic processes in ocean temperature products key to a realistic presentation of positive Indian Ocean dipole nonlinearity. Geophys. Res. Lett., 47, e2020GL089396, https://doi.org/10.1029/2020GL089396.
Zheng, X.-T., S.-P. Xie, Y. Du, L. Liu, G. Huang, and Q. Liu, 2013: Indian Ocean dipole response to global warming in the CMIP5 multimodel ensemble. J. Climate, 26, 6067–6080, https://doi.org/10.1175/JCLI-D-12-00638.1.
