Relationships of Interannual Variability Between the Equatorial Pacific and Tropical Indian Ocean in 17 CMIP5 Models

LIU Qinyu*, GUO Feiyan, and ZHENG Xiao-Tong

?

Relationships of Interannual Variability Between the Equatorial Pacific and Tropical Indian Ocean in 17 CMIP5 Models

LIU Qinyu, GUO Feiyan, and ZHENG Xiao-Tong

,,266100,

Seventeen coupled general circulation models from the Coupled Model Intercomparison Project Phase 5 (CMIP5) are employed to assess the relationships of interannual variations of sea surface temperature (SST) between the tropical Pacific (TP) and tropical Indian Ocean (TIO). The eastern/central equatorial Pacific features the strongest SST interannual variability in the models except for the model CSIRO-Mk3-6-0, and the simulated maximum and minimum are produced by models GFDL-ESM2M and GISS-E2-H respectively. However, It remains a challenge for these models to simulate the correct climate mean SST with the warm pool-cold tongue structure in the equatorial Pacific. Almost all models reproduce El Ni?o-Southern Oscillation (ENSO), Indian Ocean Dipole mode (IOD) and Indian Ocean Basin-wide mode (IOB) together with their seasonal phase lock features being simulated; but the relationship between the ENSO and IOD is different for different models. Consistent with the observation, an Indian Ocean basin-wide warming (cooling) takes place over the tropical Indian Ocean in the spring following an El Ni?o (La Ni?a) in almost all the models. In some models (, GFDL-ESM2G and MIROC5), positive ENSO and IOB events are stronger than the negative events as shown in the observation. However, this asymmetry is reversed in some other models (, HadGEM2-CC and HadGEM2-ES).

Coupled Model Intercomparison Project Phase 5; sea surface temperature; El Ni?o-Southern Oscillation; tropical Indian Ocean; tropical Pacific Ocean; interannual variability

1 Introduction

The El Ni?o-Southern Oscillation (ENSO) events in the equatorial Pacific exert a strong influence on the Indian Ocean climate (Schott, 2009). In addition to the atmospheric convection anomaly over the central equatorial Pacific, convection anomalies also appear in both the tropical West Pacific-East Indian Ocean (EIO) and the western tropical Indian Ocean (TIO) during El Ni?o (La Ni?a) period, caused by the Walker circulation anomaly. Observational and modeling studies show a clear link between SST anomalies in the tropical Pacific (TP) and Indian Ocean in boreal winter and spring (Alexander, 2002).

During the El Ni?o (La Ni?a) developing period, corresponding to the Walker circulation anomaly, anomalous easterlies (westerly) appear near the equator in the EIO. The equatorial winds reverse from westerly to easterly during the peak phase of the positive Indian Ocean Dipole (IOD) events, which locks to boreal summer and fall distinguished as a dipole in the SST anomalies (Saji,1999; Zheng, 2010). There are two different viewpoints about the relation between the IOD and ENSO: one is that the IOD is independent on the ENSO (Saji, 1999); the other is that the IOD and ENSO episodes are significantly correlated, particularly during the IOD mature phase (September–November); moreover, several recent IOD and ENSO events occurred simultaneously, most prominently in 1997 (Scott, 2009; Yang, 2010). We believe there are some independent ocean-atmosphere interaction processes in the TP and TIO during the IOD formation period, but the atmosphere bridge can set up the interconnection in the interannual variation of the SST between the TP and the TIO.

The Tropical Indian Ocean Basin-wide warms (cools), peaks in the spring of El Ni?o (La Ni?a) (Klein, 1999; Lau and Nath, 2003; Schott, 2009), and persists through boreal summer (Yang, 2007; Du, 2009; Yang, 2010) because of the downwelling ocean Rossby waves in the southwestern basin (Xie, 2002) and the heat flux changes elsewhere (Klein, 1999). This basin-wide warming (cooling) phenomenon in the TIO displays as the first empirical orthogonal function of the SST anomaly in the TIO (IOB) both in observations and model simulations (Saji, 2006; Deser, 2010; Zheng, 2011). Several previous studies suggested that IOB behaves like a capacitor anchoring summer atmospheric anomalies over the Indian–western Pacific Oceans and East Asia (Yang, 2007; Xie, 2009).

In a word, according to previous research results, there are two possible relationships of the SST interannual variation between the TP and the TIO: one is the relation between the ENSO and IOD from late summer to fall during the El Ni?o (La Ni?a) developing phase; the other is lagging correlation of the IOB in spring with ENSO during the El Ni?o (La Ni?a) decay phase.

Until longer and denser observations become available, coupled general circulation models (CGCMs) will necessarily remain as the main tools to study the relationship between ENSO and interannual variability of the TIO. The first analysis of ENSO using Coupled Model Intercomparison Project Phase 5 (CMIP5) model simulations has been made by Xueli Wang (http://www.knmi.nl/~wang/ENSO.html). Compared to the CMIP3 models, the pre-industrial simulations of the CMIP5 models are found to (1) better simulate the observed spatial patterns of the two types of ENSO and (2) have a significantly smaller inter-model diversity in ENSO intensities (Kim and Yu, 2012). The response of the IOD mode to global warming has been investigated based on CMIP5 simulations and it was proved again that the interannual variance of the IOD mode remains largely unchanged in SST, because the atmospheric feedback and zonal wind variance weaken under global warming (Zheng, 2013). A better understanding of this relationship between the TP and the TIO in the CMIP5 models will help study these phenomena and improve models to simulate these relationships well.

The present study assesses 17 CGCMs from CMIP5, first simultaneously focusing on the relationship between the interannual variations of SST in the TP and the TIO. We want to address the following questions: Can the CMIP5 models simulate the relationship of the SST interannual variability between TP and TIO? How do the CMIP5 models compare with observations? Are these relationships model-dependent or not?

The rest of the paper is organized as follows. Section 2 briefly describes the data source. Section 3 examines the simulations about climatologic mean SST, the interannual variability in TP and TIO in 17 coupled models and seasonality of ENSO, IOD and IOB of SST variability. Section 4 investigates the relationships between ENSO and IOD (IOB). Section 5 contains the discussion and summary.

2 Data and Method

The dataset used in this study includes the observation SST data (HadISST1) and the outputs of 17 models from the CMIP5 datasets. There are many experiments using the CMIP5 models (Taylor, 2012); here we particularly examine the historical experiments (http://cmip-pcmdi.llnl.gov/cmip5/). Most of the models are labeled by the name of the institution that performed the run and supplied the data. The historical experiments are simulations of the recent past (1850–2005). The imposed changing conditions (consistent with observations) include: atmospheric composition (including CO), due to both anthropogenic and volcanic influences; solar forcing; emissions or concentrations of short-lived species and natural and anthropogenic aerosols or their precursors; land use. The results in this analysis are only for the period between January 1951 and December 2005. The spatial resolution varies among models and within the same model for atmospheric and oceanic variables. To facilitate comparisons between models and with observations, we interpolated variables onto a 1?×1? latitude-longitude grid.

For comparison of the model simulations with the observations, we also analyzed observed and reanalyzed datasets. The observed SST is from the Met Office Hadley Centre’s sea ice and sea surface temperature dataset (HadISST1), which is a unique combination of monthly global fields of SST and sea ice concentration on a 1?×1? latitude-longitude grid.

3 Simulation of SST in the TP and TIO

3.1ClimatologyandInterannualVariationoftheSST

This section examines the reproducibility of climatology and interannual variation of SST over the TP and TIO in observation and the CMIP5 models. We used the last 55 years (1951–2005) of the historical model datasets to analyze the simulation of climatology mean and interannual variation. Each CMIP5 model is defined with an alphabet from (b) to (r) in Table 1. Fig.1 shows the climatologic mean and the Interannual Standard Deviation (ISD) of the SST from the observation and ensemble mean of the 17 coupled models. Compared with the observation (Fig.1a), the Pacific warm pool (SST>28℃) is simulated well (Fig.1b), but the equatorial cold tongues extends too westward and it is narrow and symmetric about the equa-tor in the ensemble mean, while it is asymmetric about the equator in the HadISST1. The area of Indian Ocean warm pool is smaller than that in the observation, especially in Bay of Bengal with 1℃ lower than the observed. There are great inter-model differences among the 17 models (Fig.2). All coupled models reproduce the Indian-Pacific warm pool, but the models (g) and (h) generate a stronger Pacific warm pool than the observed; in other 15 models the simulated equatorial cold tongues extend too westward. In the observation, the warm pool in the TIO can cover the equatorial region east of about 50?E, the Bay of Bengal and the eastern Arabian Sea. Compare to the observation, it is obvious that the areas of Indian Ocean warm pool are smaller in most of models, especially for models (e), (j) and (q), in which the Indian Ocean warm pool nearly disappear, while excessively larger Indian Ocean warm pools are simulated in (g), (h), (i) and (l) model.

Table 1 List of the observation data (HadISST) and 17 CMIP5 models

The maximum of SST ISD appears in the equatoral eastern or cental Pacific (cold tongue area) in the observation and multi-model ensemble mean (Fig.1); all the 17 CMIP5 models can capture this feature (Fig.2). Nevertheless, the maximum ISD area in model (d) is in the west of 175?E because of the cold tongue extending too westward, and the inter-model difference of the maximum ISD of SST anomaly is about 1℃ (Fig.2). The ISD in TIO for the multi-model mean is nearly 0.2℃ higher than that in observation, especially in the equatorial southeastern Indian Ocean (Fig.1). The inter-model differences of the ISD of SST anomaly in the TP and TIO are evident.

Fig.2 Magnitudes of annual mean of SST (black contours in ℃) and their interannual standard deviations (color shaded) during 1951–2005. (a) the observation of HadISST data set; (b)–(r) 17 models CMIP5 model.

3.2 Seasonal Phase Lock of the ENSO, IOD and IOB

As we know that the ENSO is the predominant inter- annual variability mode, and the IOB and IOD are the two major modes of interannual variation in TIO; The ENSOs peak in boreal winter, the IODs peak in boreal fall and the IOBs peak in the boreal spring. Whether the CMIP5 models can reproduce the realistic seasonality of these three modes of SST variability or not? Fig.3 shows the monthly SST standard deviation of ENSO, IOD and IOB indexes for the observation and multi-model ensemble mean. The ENSO index is the SST anomaly in the region of 172?E–120?W, 5?S–5?N (Alexander, 2002); the IOD index (Saji, 1999) is the difference of SST anomaly between the tropical western Indian Ocean (50?E–70?E, 10?S–10?N) and the tropical south-eastern Indian Ocean (90?E–110?E, 10?S–Equator); the IOB index is the SST anomaly in the region of 40?E–100?E, 20?S–20?N (Zheng, 2011). The ENSO indexes (Fig.3a) obviously show the seasonal phase lock with peak in winter (November-December-January) both in the observation and multi-model mean, and it is more striking in the observation; The IOD index reaches its peak during fall both in the observation and multi-model mean, but the standard deviation of the IOD index in fall is larger in models than it is in observation; There are two peaks (one in winter and the other in spring) for the IOB index both in the observation and multi-model ensemble mean. The difference is that the spring peak is larger than the winter peak in multi-model ensemble mean, which is opposite to the case of observation. All the three indexes significantly exhibit the seasonal phase lock feature as it in the observation.

In order to quantitatively present the inter-model differences of the SST ISD in TP and TIO, the ISD of different indexes from observation and 17 models are shown in Table 2. We defined two indexes TP1 and TP2: TP1 is the ENSO index during the El Ni?o (La Ni?a) developing period (ASON), and TP2 is that during the mature period (NDJ). In order to present the SST ISD in TIO, the IOD index (in ASON) and IOB index (in FMA) are also listed in Table 2. It is shown that the maximum ISD of all indexes appears in model (f) and larger than that in the observation, and the minimum ISD of all indexes appears in the model (g); In simulations of the most models, the ISDs of all indexes are close to the observed values, especially in simulations of models (e) and (k).

Fig.3 Monthly standard deviation of ENSO index (a); IOD index (b); IOB index (c) (gray bar based on HadISST and black bar based on 17 models ensemble mean SST).

Table 2 Standard deviations of four indexes based on observation data (HadISST) and 17 CMIP5 models

In summary, the models perform generally well in the simulations of climatology and interannual variability of the SST in the Tropical Pacific, the simulated spatial patterns of the ENSO are similar to the observed. However, simulating the correct warm pool-cold tongue structure in the equatorial Pacific is still a challenge for coupled models. The multi-model ensemble mean well simulates the seasonal phase lock feature for the ENSO, IOD and IOB.

4 Relationships Between the ENSO and Interannual Variation of SST in the TIO

This section investigates the relationships of SST interannual variation in the TP with that in the TIO in the observation and 17 CMIP5 models. There could be two kinds of possible relationships between ENSO and the SST inter-annual variation in the TIO. One is between ENSO and IOD: the SST anomaly is positive (negative) in west TIO and negative (positive) in East TIO during El Ni?o (La Ni?a) development. The other is between ENSO and IOB: lagging positive (negative) correlation appears in the TIO during El Ni?o (La Ni?a) decaying phase from winter to the following summer when the SST anomaly in the TIO is IOB. We will examine these two relationships.

4.1 Relationship Between ENSO and IOD

We will use the TP1 as an index of the ENSO during El Ni?o (La Ni?o) developing phase, the relationships be- tween ENSO and IOD in the observation and 17 models being shown in Figs.4 and 5. It is found that except for (i) and (l) modeland being consistent with the observation, the positive correlation coefficients between ENSO index and SST anomalies appear in the Western TIO during the ASON in all models, and are more obvious in the Northwest Indian Ocean than in the Southwest Indian Ocean. Although only eight models (b, c, d, e, o, p, q, r) have simulated the negative correlation in the EIO, in other nine models the correlation coefficients are near zero in the EIO. This result shows that during the El Ni?o (La Ni?a) developing period, the positive (negative) SST anomaly in the West TIO can be similated well in 15 models except models (i) and (l). Only in model (l), the similated SST anomaly pattern shows a south-north anti-phase mode rather than an east-west anti-phase dipole mode. From the correlation discussed above, we do not know whether the IOD appears during the El Ni?o (La Ni?a) developing period or not.

A scatter diagram is used to check the relationship between ENSO and IOD in an another way. We use the standardized IOD index (Saji, 1999) in ASON as the-coordinate, and the standardized TP1 index in ASON as the-coordinate of each star in Fig.5. In eight models (b, c, e, f, h, o, p, r), the ENSO and IOD almost simultaneously occur in the same fall, which is well consistent with the observation: in most years with positive (negative) TP1 index>0.5 (

Fig.4 Correlation coefficients of the SST anomalies averaged over the equatorial Pacific (172?E–120?W, 5?S–5?N) with the simultaneous SST anomalies at each point of the tropical Pacific and Indian Ocean in ASON for the period 1951–2005. (a) Observation; (b)–(r) 17 models (shaded colors for >90% significance).

Fig.5 Scatter diagrams for the relationships between TP1 index (X-axis) and IOD index (Y-axis) in ASON: (a) Observation; (b)–(r) 17 models. In each panel the number is the slope of each fitted line (red star: TP1 index larger than 0.5; blue star: TP1 index smaller than ?0.5; black star: the rest of cases).

4.2 Relationship Between ENSO and IOB

We use the TP2 as an index of ENSO during the El Ni?o (La Ni?a) mature period; the correlations of this index with lagged SST anomaly in each point in the TP and TIO during the sequent spring (FMA) are shown in Fig.6. It is obvious that ENSO influences the evolution of SST anomalies outside the equatorial Pacific including the TIO. In the observation (Fig.6a), a basin-wide warm-ing (cooling) takes place over the TIO following the mature phase of the El Ni?o (La Ni?a), with positive (negative) SST anomalies in the TIO lagging the SST anomalies in the central Pacific by about 3 months, which is similar to Fig.2 from Alexander(2002) (Fig.6).In the 17 CMIP5 models, the delay positive correlations between the IOB and ENSO have been reproduced, except for model (r) (Fig.6), in which the delayed negative correlation appears in the eastern part of TIO.

Fig.6 Correlation coefficients of the TP2 index with SST anomalies at each point of the tropical Pacific and Indian Ocean in the subsequent FMA for the period 1951–2005. (a) Observation; (b)–(r) 17 models (shaded colors for >90% significance).

Fig.7 Scatter diagrams for the relationships between TP2 index (X-axis) and IOB index (Y-axis) in subsequent FMA: (a) Observation; (b)–(r) 17 models. In each panel the number indicates the slope of each fitted line (red star: TP2 index larger than 0.5; blue star: TP2 index smaller than ?0.5; black star: the rest of cases).

To further explore the relationship between ENSO and IOB, we plot a scatter diagram for observation and each model. Fig.7 shows the relation between standardized TP2 index (-axis) and the lagged standardized SST anomaly in the subsequent FMA of the TIO (40?E–100?E, 20?S–20?N) (-axis in Fig.7). Consistent with the observation, all the models exhibit a lagging positive correlation between the ENSO and IOB, and the slope of each fitted line is almost the same in the 15 models as that in the observation, except for models (l) and (n) (Fig.7). Then following an El Ni?o (La Ni?a) event, an Indian Ocean basin-wide warming (cooling) takes place over the tropical Indian Ocean, peaks in the spring for almost all models.

In Fig.7, compared to the observation, the asymmetry features between the El Ni?o and La Ni?a (the absolute of positive SST anomaly is larger than the absolute of negative SST anomaly (Hong, 2010)) are different among the 17 models: In almost all the models the similar asymmetry feature is consistent with the observation; in models (e) and (p), the asymmetry features exceed that in the observation. On the contrary, in the models (c), (j) and (r), there is an opposite asymmetry between El Ni?o and La Ni?a (the absolute of the positive SST anomaly is smaller than absolute of negative SST anomaly). This similar asymmetry also appears in the IOB during the subsequent FMA, because the relationship between ENSO and IOB is very close. In summary, the relationship between ENSO and IOB has been simulated well in the 17 CMIP models.

5 Summary and Discussion

We have examined the relationship of SST interannual variability between the TP and the TIO from 17 CMIP5 models. As in the observation, the ENSO is the dominant mode of the SST anomaly in the TP. The 17 models per- form generally better in simulating the SST interannual variability in the TP; but it is still a challenge for those 17 models to simulate the realistic climatology SST with the warm pool-cold tongue structure in the equatorial Pacific. The models well simulates the seasonal phase lock feature for the ENSO, IOD and IOB.

From the above correlation analyses we reach the fol- lowing conclusions: in the 17 models the positive correlation of the SST anomaly between the TP and Western TIO are well consistent with the observation, but the negative correlations of SST anomaly between EIO and ENSO in most of the models do not agree with the observation. Although this inter-model discrepancy does not affect on the existence of the IOD in almost all the models’ simulations, the relation between the ENSO and IOD shows diversity of results among these models. It is again illustrated that the IOD is independent on the ENSO. Compared to the observation, the relationship between ENSO and IOB is simulated well in the 17 CMIP models. In the GFDL-ESM2M model the simulated amplitudes of the ENSO and IOB are enlarged, but in GISS-E2-H model the simulated amplitudes of the ENSO and IOB are the smallest in the 17 models. In some models (GFDL-ESM2G and MIROC5), positive ENSO and IOB events are stronger than the negative events as shown in the observation. However, this asymmetry is reversed in some other models (HadGEM2-CC and HadGEM2-ES).

Acknowledgements

This work is supported by the National Basic Research Program of China 2012CB955602 and 2012CB955603 and the Natural Science Foundation of China (41176006, 40921004 and 41106010). We thank the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and the climate modeling groups (listed in Table 1 of this paper) for producing and making available their model output.

Alexander, M. A., Bladé, I., Newman, M., Lanzante, J. R., Lau, N. C., and Scott, J. D., 2002. The atmospheric bridge: The influence of ENSO teleconnections on air-sea interaction over the global oceans., 15: 2205-2231, DOI: http://dx.doi.org/10.1175/1520-0442.

Deser, C., Alexander, M. A., Xie, S.P., and Phillips, A. S., 2010. Sea surface temperature variability: Patterns and mechanisms., 2: 115-143, DOI:10.1146/annurev-marine-120408-151453.

Du, Y., Xie, S. P., Huang, G., and Hu, K., 2009. Role of air-sea interaction in the long persistence of El Ni?o-induced North Indian Ocean warming., 22: 2023-2038.

Hong, C. C., Li, T., Lin, H., and Chen, Y. C., 2010. Asymmetry of the Indian Ocean basinwide SST anomalies: Roles of ENSO and IOD., 23: 3563-3576.

Kim, S. T., and Yu, J. Y., 2012. The two types of ENSO in CMIP5 models., 39, L11704, DOI: 10.1029/2012GL052006.

Klein, S. A., Soden, B. J., and Lau, N. C., 1999. Remote sea surface temperature variations during ENSO: Evidence for a tropical atmospheric bridge., 12: 917-932.

Lau, N. C., and Nath, M. J., 2003. Atmosphere–ocean variations in the Indo-Pacific sector during ENSO episodes., 16: 3-20.

Saji, N. H., Goswami, B. N., Vinayachandran, P. N., and Yamagata, T., 1999. A dipole mode in the tropical Indian Ocean., 401: 360-363.

Saji, N. H., Xie, S. P., and Yamagata, T., 2006. Tropical Indian Ocean variability in the IPCC 20th-century climate simulations., 19: 4397-4417.

Schott, F. A., Xie, S. P., and McCreary, J. P., 2009. Indian Ocean circulation and climate variability.,47, RG1002, DOI: 10.1029/2007RG000245.

Taylor, K. E., Stouffer, R. J., and Meehl, G. A., 2012. An overview of CMIP5 and the experiment design., 93: 485-498.

Xie, S. P., Annamalai, H., Schott, F. A., and McCreary, J. P., 2002. Structure and mechanisms of South Indian Ocean climate variability., 15: 864-878.

Xie, S. P., Hu, K., Hafner, J., Tokinaga, H., Du, Y., Huang, G., and Sampe, T., 2009. Indian Ocean capacitor effect on Indo-western Pacific climate during the summer following El Ni?o., 22: 730-747.

Yang, J. L., Liu, Q. Y., and Liu, Z., 2010. Linking observations of the Asian monsoon to the Indian Ocean SST: Possible roles of Indian Ocean basin mode and dipole mode., 23: 5889-5902.

Yang, J., Liu, Q., Xie, S. P., Liu, Z., and Wu, L., 2007. Impact of the Indian Ocean SST basin mode on the Asian summer monsoon., 34, L02708, DOI: 10.1029/2006GL028571.

Zheng, X. T., Xie, S. P., and Liu, Q. Y., 2011. Response of the Indian Ocean basin mode and its capacitor effect to global warming.,24 (23): 6146-6164. DOI: 10.1175/2011JCLI4169.1.

Zheng, X. T., Xie, S. P., Du, Y., Liu, L., Huang, G., and Liu, Q., 2013. Indian Ocean Dipole response to global warming in the CMIP5 multi-model ensemble., DOI: 10.1175/JCLI-D-12-00638.1, in press.

Zheng, X. T., Xie, S. P., Vecchi, G. A., Liu, Q. Y., and Hafner, J., 2010. Indian Ocean Dipole response to global warming: Analysis of ocean–atmospheric feedbacks in a coupled model., 23: 1240-1253.

(Edited by Xie Jun)

10.1007/s11802-013-2195-8

ISSN 1672-5182, 2013 12 (2): 237-244

. Tel: 0086-532-66782556 E-mail: liuqy@ouc.edu.cn

(October 29, 2012; revised January 4, 2013; accepted January 30, 2013)

? Ocean University of China, Science Press and Springer-Verlag Berlin Heidelberg 2013