• Annamalai, H., P. Liu, and S.-P. Xie, 2005: Southwest Indian Ocean SST variability: Its local effect and remote influence on Asian monsoons. J. Climate, 18, 41504167, https://doi.org/10.1175/JCLI3533.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ashok, K., W. L. Chan, T. Motoi, and T. Yamagata, 2004: Decadal variability of the Indian Ocean dipole. Geophys. Res. Lett., 31, L24207, https://doi.org/10.1029/2004GL021345.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Atlas, R., R. N. Hoffman, J. Ardizzone, S. M. Leidner, J. C. Jusem, D. K. Smith, and D. Gombos, 2011: A cross-calibrated, multiplatform ocean surface wind velocity product for meteorological and oceanographic applications. Bull. Amer. Meteor. Soc., 92, 157174, https://doi.org/10.1175/2010BAMS2946.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Balmaseda, M. A., K. E. Trenberth, and E. Källén, 2013: Distinctive climate signals in reanalysis of global ocean heat content. Geophys. Res. Lett., 40, 17541759, https://doi.org/10.1002/grl.50382.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Baquero-Bernal, A., and M. Latif, 2005: Wind-driven oceanic Rossby waves in the tropical South Indian Ocean with and without an active ENSO. J. Phys. Oceanogr., 35, 729746, https://doi.org/10.1175/JPO2723.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Birol, F., and R. Morrow, 2001: Source of the baroclinic waves in the southeast Indian Ocean. J. Geophys. Res., 106, 91459160, https://doi.org/10.1029/2000JC900044.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cai, W., and et al. , 2018: Increased variability of eastern Pacific El Niño under greenhouse warming. Nature, 564, 201206, https://doi.org/10.1038/s41586-018-0776-9.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chambers, D. P., B. D. Tapley, and R. H. Stewart, 1999: Anomalous warming in the Indian Ocean coincident with El Niño. J. Geophys. Res., 104, 30353047, https://doi.org/10.1029/1998JC900085.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chelliah, M., and P. Arkin, 1992: Large-scale interannual variability of monthly outgoing longwave radiation anomalies in the global tropics. J. Climate, 5, 371389, https://doi.org/10.1175/1520-0442(1992)005<0371:LSIVOM>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, G., W. Han, Y. Li, D. Wang, and M. McPhaden, 2015: Seasonal-to-interannual time scale dynamics of the Equatorial Undercurrent in the Indian Ocean. J. Phys. Oceanogr., 45, 15321553, https://doi.org/10.1175/JPO-D-14-0225.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, G., W. Han, Y. Li, and D. Wang, 2016: Interannual variability of equatorial eastern Indian Ocean upwelling: Local versus remote forcing. J. Phys. Oceanogr., 46, 789807, https://doi.org/10.1175/JPO-D-15-0117.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dee, D., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, https://doi.org/10.1002/qj.828.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Deepa, J. S., C. Gnanaseelan, R. Kakatkar, A. Parekh, and J. S. Chowdary, 2018: The interannual sea level variability in the Indian Ocean as simulated by an ocean general circulation model. Int. J. Climatol., 38, 11321144, https://doi.org/10.1002/joc.5228.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ebisuzaki, W., 1997: A method to estimate the statistical significance of a correlation when the data are serially correlated. J. Climate, 10, 21472153, https://doi.org/10.1175/1520-0442(1997)010<2147:AMTETS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fu, L. L., and R. D. Smith, 1996: Global ocean circulation from satellite altimetry and high resolution computer simulation. Bull. Amer. Meteor. Soc., 77, 26252636, https://doi.org/10.1175/1520-0477(1996)077<2625:GOCFSA>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fukumori, I., R. Raghunath, and L.-L. Fu, 1998: Nature of global large-scale sea level variability in relation to the atmospheric forcing: A modeling study. J. Geophys. Res., 103, 54935512, https://doi.org/10.1029/97JC02907.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gill, E. A., 1982: Atmosphere–Ocean Dynamics. Academic Press, 662 pp.

  • Gnanaseelan, C., and B. H. Vaid, 2010: Interannual variability in the biannual Rossby waves in the tropical Indian Ocean and its relation to Indian Ocean dipole and El Niño forcing. Climate Dyn., 60, 2740, https://doi.org/10.1007/s10236-009-0236-z.

    • Search Google Scholar
    • Export Citation
  • Han, W., 2005: Origins and dynamics of the 90-day and 30-60-day variations in the equatorial Indian Ocean. J. Phys. Oceanogr., 35, 708728, https://doi.org/10.1175/JPO2725.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., and P. J. Webster, 2002: Forcing mechanisms of sea-level interannual variability in the Bay of Bengal. J. Phys. Oceanogr., 32, 216239, https://doi.org/10.1175/1520-0485(2002)032<0216:FMOSLI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., J. P. McCreary, D. L. T. Anderson, and A. J. Mariano, 1999: Dynamics of the eastward surface jets in the equatorial Indian Ocean. J. Phys. Oceanogr., 29, 21912209, https://doi.org/10.1175/1520-0485(1999)029<2191:DOTESJ>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., P. J. Webster, R. Lukas, P. Hacker, and A. Hu, 2004: Impact of atmospheric intraseasonal variability in the Indian Ocean: Low-frequency rectification in equatorial surface current and transport. J. Phys. Oceanogr., 34, 13501372, https://doi.org/10.1175/1520-0485(2004)034<1350:IOAIVI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., J. P. McCreary, Y. Masumoto, J. Vialard, and B. Duncan, 2011: Basin resonances in the equatorial Indian Ocean. J. Phys. Oceanogr., 41, 12521270, https://doi.org/10.1175/2011JPO4591.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., G. Meehl, D. Stammer, A. Hu, B. Hamlington, J. Kenigson, H. Palanisamy, and P. Thompson, 2017a: Spatial patterns of sea level variability associated with natural internal climate modes. Surv. Geophys., 38, 217250, https://doi.org/10.1007/s10712-016-9386-y.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., G. Meehl, A. Hu, J. Zheng, J. Kenigson, J. Vialard, B. Rajagopalan, and Yanto, 2017b: Decadal variability of Indian and Pacific Walker cells: Do they co-vary on decadal timescales? J. Climate, 30, 84478468, https://doi.org/10.1175/JCLI-D-16-0783.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., D. Stammer, G. A. Meehl, A. Hu, F. Sienz, and L. Zhang, 2018: Multi-decadal trend and decadal variability of the regional sea level over the Indian Ocean since the 1960s: Roles of climate modes and external forcing. Climate, 6, 51, https://doi.org/10.3390/cli6020051.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hermes, J. C., and C. J. C. Reason, 2008: Annual cycle of the South Indian Ocean (Seychelles-Chagos) thermocline ridge in a regional ocean model. J. Geophys. Res., 113, C04035, https://doi.org/10.1029/2007JC004363.

    • Search Google Scholar
    • Export Citation
  • Hosoda, S., T. Ohira, and T. Nakamura, 2008: A monthly mean dataset of global oceanic temperature and salinity derived from Argo float observations. JAMSTEC Rep. Res. Dev., 8, 4759, https://doi.org/10.5918/jamstecr.8.47.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hu, S., and et al. , 2019: Interannual to decadal variability of upper-ocean salinity in the southern Indian Ocean and the Role of the Indonesian Throughflow. J. Climate, 32, 64036421, https://doi.org/10.1175/JCLI-D-19-0056.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huang, B., and J. L. Kinter III, 2002: Interannual variability in the tropical Indian Ocean. J. Geophys. Res., 107, 3199, https://doi.org/10.1029/2001JC001278.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Izumo, T., and et al. , 2010: Influence of the state of the Indian Ocean Dipole on the following year’s El Niño. Nat. Geosci., 3, 168172, https://doi.org/10.1038/ngeo760.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Jury, M. R., and B. Huang, 2004: The Rossby wave as a key mechanism of Indian Ocean climate variability. Deep-Sea Res. I, 51, 21232136, https://doi.org/10.1016/j.dsr.2004.06.005.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Klein, S. A., B. J. Soden, and N.-C. Lau, 1999: Remote sea surface variations during ENSO: Evidence for a tropical atmospheric bridge. J. Climate, 12, 917932, https://doi.org/10.1175/1520-0442(1999)012<0917:RSSTVD>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Krishnaswamy, J., S. Vaidyanathan, B. Rajagopalan, M. Bonell, M. Sankaran, R. S. Bhalla, and S. Badiger, 2015: Non-stationary and non-linear influence of ENSO and Indian Ocean dipole on the variability of Indian monsoon rainfall and extreme rain events. Climate Dyn., 45, 175184, https://doi.org/10.1007/s00382-014-2288-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kumar, K. K., B. Rajagopalan, and M. A. Cane, 1999: On the weakening relationship between the Indian monsoon and ENSO. Science, 284, 21562159, https://doi.org/10.1126/science.284.5423.2156.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kumar, K. K., B. Rajagopalan, M. Hoerling, G. Bates, and M. Cane, 2006: Unraveling the mystery of Indian monsoon failure during El Niño events. Science, 314, 115119, https://doi.org/10.1126/science.1131152.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Levitus, S., and et al. , 2012: World ocean heat content and thermosteric sea level change (0–2000 m), 1955–2010. Geophys. Res. Lett., 39, L10603, https://doi.org/10.1029/2012GL051106.

    • Search Google Scholar
    • Export Citation
  • Li, Y., and W. Han, 2015: Decadal sea level variations in the Indian Ocean investigated with HYCOM: Roles of climate modes, ocean internal variability and stochastic wind forcing. J. Climate, 28, 91439165, https://doi.org/10.1175/JCLI-D-15-0252.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Masumoto, Y., and G. Meyers, 1998: Forced Rossby waves in the southern tropical Indian Ocean. J. Geophys. Res., 103, 27 58927 602, https://doi.org/10.1029/98JC02546.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., 1981: A linear stratified ocean model of the coastal undercurrent. Philos. Trans. Roy. Soc. London, 302A, 385413, https://doi.org/10.1098/rsta.1981.0176.

    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., P. K. Kundu, and R. L. Molinari, 1993: A numerical investigation of dynamics, thermodynamics and mixed-layer processes in the Indian Ocean. Prog. Oceanogr., 31, 181244, https://doi.org/10.1016/0079-6611(93)90002-U.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., W. Han, D. Shankar, and S. R. Shetye, 1996: Dynamics of the East India Coastal Current: 2. Numerical solutions. J. Geophys. Res., 101, 13 99314 010, https://doi.org/10.1029/96JC00560.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Murtugudde, R., S. Signorini, J. Christian, A. Busalacchi, C. McClain, and J. Picaut, 1999: Ocean color variability of the tropical Indo-Pacific basin observed by SeaWiFS during 1997-98. J. Geophys. Res., 104, 18 35118 366, https://doi.org/10.1029/1999JC900135.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Murtugudde, R., J. P. McCreary, and A. J. Busalacchi, 2000: Oceanic processes associated with anomalous events in the Indian Ocean with relevance to 1997–1998. J. Geophys. Res., 105, 32953306, https://doi.org/10.1029/1999JC900294.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nagura, M., and M. J. McPhaden, 2010: Dynamics of zonal current variations associated with the Indian Ocean dipole. J. Geophys. Res., 115, C11026, https://doi.org/10.1029/2010JC006423.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Paek, H., J.-Y. Yu, and C. Qian, 2017: Why were the 2015/2016 and 1997/1998 extreme El Niños different? Geophys. Res. Lett., 44, 18481856, https://doi.org/10.1002/2016GL071515.

    • Search Google Scholar
    • Export Citation
  • Périgaud, C., and P. Delecluse, 1992: Annual sea level variations in the southern tropical Indian Ocean from Geosat and shallow-water simulations. J. Geophys. Res., 97, 20 16920 178, https://doi.org/10.1029/92JC01961.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Périgaud, C., and P. Delecluse, 1993: Interannual sea level variations in the tropical Indian Ocean from Geosat and shallow water simulations. J. Phys. Oceanogr., 23, 19161934, https://doi.org/10.1175/1520-0485(1993)023<1916:ISLVIT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Petris, G., 2010: dlm: Bayesian and likelihood analysis of dynamic linear models. R package version 1.1-1, https://rdrr.io/cran/dlm/.

  • Petris, G., S. Petrone, and P. Campagnoli, 2009: Dynamic Linear Models with R. Springer, 252 pp (see especially pp. 31–84).

    • Crossref
    • Export Citation
  • Potemra, J. T., 2001: Contribution of equatorial Pacific winds to southern tropical Indian Ocean Rossby waves. J. Geophys. Res., 106, 24072422, https://doi.org/10.1029/1999JC000031.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rao, S. A., and S. K. Behera, 2005: Subsurface influence on SST in the tropical Indian Ocean: Structure and interannual variability. Dyn. Atmos. Oceans, 39, 103135, https://doi.org/10.1016/j.dynatmoce.2004.10.014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, https://doi.org/10.1029/2002JD002670.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • R Development Core Team, 2016: R: A language and environment for statistical computing. R Foundation for Statistical Computing, https://www.R-project.org/.

  • Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360363, https://doi.org/10.1038/43854.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schott, F. A., S.-P. Xie, and J. P. McCreary, 2009: Indian Ocean circulation and climate variability. Rev. Geophys., 47, RG1002, https://doi.org/10.1029/2007RG000245.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shinoda, T., H. H. Hendon, and M. A. Alexander, 2004: Surface and subsurface dipole variability in the Indian Ocean and its relation with ENSO. Deep-Sea Res. I, 51, 619635, https://doi.org/10.1016/j.dsr.2004.01.005.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Susanto, R. D., A. L. Gordon, and Q. N. Zheng, 2001: Upwelling along the coasts of Java and Sumatra and its relation to ENSO. Geophys. Res. Lett., 28, 15991602, https://doi.org/10.1029/2000GL011844.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Takahashi, K., A. Montecinos, K. Goubanova, and B. Dewitte, 2011: ENSO regimes: Reinterpreting the canonical and Modoki El Niño. Geophys. Res. Lett., 38, L10704, https://doi.org/10.1029/2011GL047364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Timmermann, A., and et al. , 2018: El Niño–Southern Oscillation complexity. Nature, 559, 535545, https://doi.org/10.1038/s41586-018-0252-6.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tozuka, T., T. Yokoi, and T. Yamagata, 2010: A modeling study of interannual variations of the Seychelles Dome. J. Geophys. Res., 115, C04005, https://doi.org/10.1029/2009JC005547.

    • Search Google Scholar
    • Export Citation
  • Trenary, L., and W. Han, 2012: Intraseasonal to interannual variability of South Indian Ocean sea level and thermocline: Remote versus local forcing. J. Phys. Oceanogr., 42, 602627, https://doi.org/10.1175/JPO-D-11-084.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wang, L., C. J. Koblinsky, and S. Howden, 2001: Annual Rossby wave in the Southern Indian Ocean: Why does it “appear” to break down in the middle ocean? J. Phys. Oceanogr., 31, 5474, https://doi.org/10.1175/1520-0485(2001)031<0054:ARWITS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 356360, https://doi.org/10.1038/43848.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., J. Scott, R. Hoffman, M. Leidner, R. Atlas, and J. Ardizzone, 2015: Remote Sensing Systems Cross-Calibrated Multi-Platform (CCMP) 6-hourly ocean vector wind analysis product on 0.25 deg grid, version 2.0. Remote Sensing Systems, accessed 2019, www.remss.com/measurements/ccmp.

  • White, W. B., 2001: Evidence for coupled Rossby waves in the annual cycle of the Indo-Pacific Ocean. J. Phys. Oceanogr., 31, 29442957, https://doi.org/10.1175/1520-0485(2001)031<2944:EFCRWI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Woodberry, K. E., M. E. Luther, and J. J. O’Brien, 1989: The wind-driven seasonal circulation in the southern tropical Indian Ocean. J. Geophys. Res., 94, 17 98518 002, https://doi.org/10.1029/JC094Ic12P17985.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xie, S.-P., and P. A. Arkin, 1996: Analysis of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9, 840858, https://doi.org/10.1175/1520-0442(1996)009<0840:AOGMPU>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xie, S.-P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78, 25392558, https://doi.org/10.1175/1520-0477(1997)078<2539:GPAYMA>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xie, S.-P., H. Annamalai, F. A. Schott, and J. P. McCreary, 2002: Structure and mechanisms of south Indian Ocean climate variability. J. Climate, 15, 864878, https://doi.org/10.1175/1520-0442(2002)015<0864:SAMOSI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 33523368, https://doi.org/10.1175/2010JCLI3429.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yang, J., L. Yu, C. J. Koblinsky, and D. Adamec, 1998: Dynamics of the seasonal variations in the Indian Ocean from TOPEX/POSEIDON sea surface height and an ocean model. Geophys. Res. Lett., 25, 19151918, https://doi.org/10.1029/98GL01401.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yokoi, T., T. Tozuka, and T. Yamagata, 2008: Seasonal variation of the Seychelles Dome. J. Climate, 21, 37403754, https://doi.org/10.1175/2008JCLI1957.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yokoi, T., T. Tozuka, and T. Yamagata, 2009: Seasonal variation of the Seychelles Dome simulated in the CMIP3 models. J. Phys. Oceanogr., 39, 449457, https://doi.org/10.1175/2008JPO3914.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yu, J.-Y., Y. Zou, S. T. Kim, and T. Lee, 2012: The changing impact of El Niño on US winter temperatures. Geophys. Res. Lett., 39, L15702, https://doi.org/10.1029/2012GL052483.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yu, W., B. Xiang, L. Liu, and N. Liu, 2005: Understanding the origins of interannual thermocline variations in the tropical Indian Ocean. Geophys. Res. Lett., 32, L24706, https://doi.org/10.1029/2005GL024327.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhuang, W., M. Feng, Y. Du, A. Schiller, and D. Wang, 2013: Low-frequency sea level variability in the southern Indian Ocean and its impacts on the oceanic meridional transports. J. Geophys. Res. Oceans, 118, 13021315, https://doi.org/10.1002/jgrc.20129.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) Annual mean, (b) January, and (c) September climatology of depth of 20°C isotherm (D20; color) and ERA-Interim surface wind stress (arrows) for the 1993–2016 period. The boxed region in the west shows the Seychelles–Chagos thermocline ridge (SCTR) region (2°–15°S, 50°–80°E), the mean upwelling zone of the tropical Indian Ocean; the boxed area in the east shows the eastern node of the Indian Ocean dipole (IOD) area (10°S–0°N, 90°–115°E), which includes the seasonal upwelling region along the coasts of Sumatra and Java during boreal summer–fall. (d) September–November (SON) averaged D20 anomaly (D20A) and wind stress anomaly for 1997, when positive IOD (pIOD) and El Niño co-occurred; (e) as in (d), but for 2012, an independent pIOD (no El Niño or La Niña); (f) as in (d), but for 2009, an El Niño year (no IOD). The D20 is based on the monthly ECMWF Ocean Reanalysis System 4 (ORAS4) data from 1993 to 2016.

  • View in gallery

    (a) Monthly Niño-3.4 index (red line) and DMI (blue line) calculated from monthly HadISST with their linear trends and monthly climatology of 1993–2016 removed. Units: °C. Their correlation coefficient and standard deviations (STD) for each time series are shown at the bottom of the panel. (b) Time series of monthly DMI (red line), same as the blue curve in (a), and independent DMI with ENSO effect removed using conventional static linear regression (blue line). In each panel, the dash–dotted horizontal red (blue) lines show ±1 STD of the corresponding index. All correlation coefficients exceed 95% significance. Note that here and elsewhere in this paper, significance estimates for correlation coefficients are based on Ebisuzaki (1997).

  • View in gallery

    Standard deviation (STD) of monthly anomalies of D20 (units: m) based on (a) ORAS4 reanalysis from 1993–2016, (b) ORAS4 from 2001–16, and (c) MOAA-GPV from 2001–16. Also shown are the STD maps of monthly (d) WOA13 upper 700-m thermosteric sea level anomaly (SLA; cm) for 1993–2015, (e) ORAS4 sea surface height anomalies (SSHA; cm) for 1993–2016, and (f) AVISO SSHA for 1993–2016. The black dashed boxes show the east Indian Ocean (EIO) and west Indian Ocean (WIO) mean upwelling zones, the same boxes as in Fig. 1. The black solid boxes show the maximum STD region based on both ORAS4 and AVISO SSHA (5°–16°S, 66° to 88°E).

  • View in gallery

    Time series of monthly (a) ORAS4 D20A, (b) ORAS4 SSHA, (c) AVISO SSHA, and (d) SSTA from HadISST averaged in the WIO (red line) and EIO (blue line) mean and seasonal upwelling zones (boxed regions of Fig. 1). In (d), the two dotted lines show the corresponding SON mean SSTA. The correlation and STD values are shown at the bottom of each panel. (e) Time series of D20A averaged over the maximum D20A region (central box of Fig. 3); (f) as in (e), but for ORAS4 and AVISO SSHA.

  • View in gallery

    Bimonthly STD maps of AVISO SSHA from 1993–2016 for (a) total SSHA, (b) SSHA due to ENSO effect (b1 × ENSO), and (c) SSHA due to IOD effect (b1 × IOD) from the Bayesian dynamical linear model (DLM). The boxes are the same as in Fig. 3.

  • View in gallery

    Time series of satellite-observed AVISO SSHA (black line), SSHA explained by ENSO using the conventional static linear model (SLM; blue line) and Bayesian DLM (red line), for the (a) EIO box, (b) WIO box, and (c) maximum SSHA box shown in Figs. 3 and 5. (d)–(f) As in (a)–(c), but for SSHA explained by the IOD. Specifically, the blue and red lines are the b1X1 terms of Eqs. (2) and (3), respectively, with X1 being the Niño-3.4 index for (a)–(c) and DMI for (d)–(f). The STD for each time series and the correlations between AVISO observed SSHA and ENSO- or IOD-explained SSHA using SLM (Bayesian DLM) are shown at the bottom of each panel. All correlation coefficients exceed 95% significance.

  • View in gallery

    The leading EOF mode (EOF1) of atmospheric and oceanic variability caused by (a)–(c) ENSO and (d)–(f) the IOD from 1993 to 2016 using Bayesian DLM. Color shading shows EOF1 of (a) SSHA, (b) precipitation anomaly, and (c) outgoing longwave radiation anomaly (OLRA) induced by ENSO. Values within parentheses show variance explained by EOF1. The EOF1s of surface wind stress (vectors) and Ekman pumping velocity (we, line contours), defined as we = Curl(τ/ρwf) = (∂/∂x)(τy/ρwf) − (∂/∂y)(τx/ρwf), are also plotted in each panel. (d)–(f) As in (a)–(c), respectively, but for IOD effect. Unit for we is 1 × 107 m s−1, and positive (negative) values are in red (green), with values of 3, 6, −24, −18, −12, and −6. The units for SSHA, precipitation anomaly, OLRA, and anomalous wind stress are cm, mm day−1, W m−2, and N m−2, respectively.

  • View in gallery

    Time series of Bayesian DLM coefficients, that is, b1 of Eq. (3), in the EIO (eastern box of Fig. 5) for (a) ENSO (red curve), and (b) the IOD (blue curve). (c),(d) As in (a),(b), but for the WIO upwelling region (western box of Fig. 5). (e),(f) As in (a),(b), but for the central box of Fig. 5 where SSHA reaches the maximum. The horizontal dashed–dotted black line in each panel shows the corresponding SLM coefficient, that is, b1 of Eq. (2), which is a constant value. The El Niño (La Niña) years are marked as yellow dots (green diamonds) in (a), (c), and (e), and “×” is added to the yellow dots for EP El Niños. Positive (negative) IOD events are marked as red dots (green diamonds) in (b), (d), and (f).

  • View in gallery

    SSTA (color shading; °C) and Ekman pumping velocity (contours) for November–January (NDJ) mean of (a) 1997 El Niño, (b) 2015 El Niño, and (c) the difference between them [(a) − (b)]. Units for we are 1 × 106 m s−1. The positive (negative) values are in solid black (dashed black) with values of 2, 12, 24, −24, −12, and −2. (d)–(f) As in (a)–(c), but for OLRA (W m−2) and surface wind stress.

  • View in gallery

    NDJ mean SSHA (color shading; cm) and surface wind stress (arrows) for the composite of (a) all EP years, (b) all CP years, and (c) their difference [(a) − (b)] for the El Niño events since 1979 (see Table 1). (d)–(f) As in (a)–(c), but for OLRA (color shading; W m−2) and surface wind stress. SSHA data are from ORAS4 reanalysis, surface wind stress data are from the ERA-Interim reanalysis, and OLR data are from satellite observations.

  • View in gallery

    As in Fig. 10, but for the composites of the pure EP and pure CP El Niño years that did not co-occur with pIOD events (see Table 1 for specific years).

  • View in gallery

    The STD of monthly SSHA (unit: cm) based on the Linear Ocean Model (LOM) (a) main run (MR), denoted by LOM_MR, (b) ENSO run (LOM_ENSO), and (c) IOD run (LOM_IOD), which are forced by the monthly CCMP2 surface wind stress from 1993 to 2016. The boxes are the same as in Figs. 3 and 5. Time series of upwelling anomalies based on AVISO SSHA (black dashed line), LOM_MR (black solid line), LOM_ENSO (red line), and LOM_IOD (blue line) averaged in the (d) EIO, (e) WIO, and (f) Max Region. The STD for each time series and the correlations between AVISO SSHA and LOM_MR, LOM_MR and LOM_ENSO, and LOM_MR and LOM_IOD are shown at the bottom of each box. For the upwelling averaged in the EIO box, the correlation between AVISO SSHA and LOM_MR is r = 0.90 (rcrit95% = 0.32), the correlation between LOM_MR and LOM_ENSO is r = 0.65 (rcrit95% = 0.34), and that between LOM_MR and LOM_IOD is r = 0.79 (rcrit95% = 0.33). For the WIO box, the correlation between AVISO SSHA and LOM_MR is r = 0.90 (rcrit95% = 0.35), the correlation between LOM_MR and LOM_ENSO is r = 0.80 (rcrit95% = 0.38), and that between LOM_MR and LOM_IOD is r = 0.65 (rcrit95% = 0.39). And for the maximum SSHA box, the correlation between AVISO SSHA and LOM_MR is r = 0.87 (rcrit95% = 0.38), the correlation between LOM_MR and LOM_ENSO is r = 0.76 (rcrit95% = 0.40), and that between LOM_MR and LOM_IOD is r = 0.65 (rcrit95% = 0.39).

  • View in gallery

    The STD of monthly SSHA (unit: cm) based on the LOM (a) main run (MR), denoted by LOM_MR, and (b) damp run (LOM_DAMP), which has a damping along the eastern Indian Ocean boundary from January 1993 to December 2016 and its results measure the effects of direct wind forcing in the ocean interior without the effect of eastern equatorial Indian Ocean boundary. The boxes are the same as in Figs. 3 and 5. (c),(d) Time series of upwelling anomalies based on LOM_MR (black line), LOM_DAMP (red line), and (LOM_MR − LOM_DAMP; green line) averaged in the EIO and WIO box, respectively. The correlation between LOM_MR and LOM_DAMP SSHA time series is shown at the bottom of each box, and the SSHA STDs from the MR and (LOM_MR − LOM_DAMP) are also shown.

  • View in gallery

    OND mean maps of SSHA based on the LOM_ENSO run for (a) 1997 El Niño, (b) 2015 El Niño, and (c) the difference between them, which is calculated by subtracting 2015 values from 1997 values. The boxes are as in Fig. 3.

  • View in gallery

    Longitude–time plot of monthly SSHA along ~5°S (5.25°–4.5°S average) for the composites of pure El Niño years (2002, 2004, 2009) based on (a) AVISO data, (b) LOM_MR, (c) LOM_ENSO, and (d) the SSHA due to ENSO effect (b1 × ENSO) from Bayesian DLM using the result of LOM_MR; to better reveal the propagation signals, 24-month SSHAs (January 2002–December 2003, January 2004–December 2005, and January 2009–December 2010) are shown. (e)–(h) (second row) As in (a)–(d), but for the 2012 pure positive IOD year, January 2012–December 2013. (i)–(l) (third row) As in (a)–(d), but for the composites of ENSO/IOD co-occuring years (1994, 1997, 2006 and 2015), and the results shown are AVISO SSHA, LOM_MR, LOM_ENSO, and b1 × ENSO, respectively, from January of the event year to December of the following year. (m)–(p) As in (i)–(l), but that (o) and (p) represent LOM_IOD and b1 × IOD.

  • View in gallery

    Time series of the monthly anomalous upwelling, that is, SSHA (cm) based on LOM_ENSO experiment (black line) and the ENSO effect from Bayesian DLM (red line) averaged in the (a) EIO, (c) WIO, (e) Max Region, and (g) boxed region of (2°–15°S, 50°–65°E) within the WIO box for the period of 1993 to 2016. (b),(d),(f),(h) As in (a),(c),(e),(g), respectively, but for the SSHA from LOM_IOD experiment (black line) and the IOD effect from Bayesian DLM (red line).

  • View in gallery

    The STD maps of monthly SSHA (cm) based on the sum of (a) the first 25 baroclinic modes, (b) the first two baroclinic modes, (c) the first three baroclinic modes, (e) mode 4 to mode 8, and (g) mode 9 to mode 25 of the LOM_MR result. (d),(f),(h) The contributions for modes 1, 2, and 3, respectively. The boxes are the same as those of Fig. 3.

All Time Past Year Past 30 Days
Abstract Views 129 129 0
Full Text Views 345 345 18
PDF Downloads 312 312 13

Effects of Climate Modes on Interannual Variability of Upwelling in the Tropical Indian Ocean

View More View Less
  • 1 Department of Atmospheric and Oceanic Sciences, University of Colorado Boulder, Boulder, Colorado
© Get Permissions
Free access

Abstract

This paper investigates interannual variability of the tropical Indian Ocean (IO) upwelling through analyzing satellite and in situ observations from 1993 to 2016 using the conventional Static Linear Regression Model (SLM) and Bayesian Dynamical Linear Model (DLM), and performing experiments using a linear ocean model. The analysis also extends back to 1979, using ocean–atmosphere reanalysis datasets. Strong interannual variability is observed over the mean upwelling zone of the Seychelles–Chagos thermocline ridge (SCTR) and in the seasonal upwelling area of the eastern tropical IO (EIO), with enhanced EIO upwelling accompanying weakened SCTR upwelling. Surface winds associated with El Niño–Southern Oscillation (ENSO) and the IO dipole (IOD) are the major drivers of upwelling variability. ENSO is more important than the IOD over the SCTR region, but they play comparable roles in the EIO. Upwelling anomalies generally intensify when positive IODs co-occur with El Niño events. For the 1979–2016 period, eastern Pacific (EP) El Niños overall have stronger impacts than central Pacific (CP) and the 2015/16 hybrid El Niño events, because EP El Niños are associated with stronger convection and surface wind anomalies over the IO; however, this relationship might change for a different interdecadal period. Rossby wave propagation has a strong impact on upwelling in the western basin, which causes errors in the SLM and DLM because neither can properly capture wave propagation. Remote forcing by equatorial winds is crucial for the EIO upwelling. While the first two baroclinic modes capture over 80%–90% of the upwelling variability, intermediate modes (3–8) are needed to fully represent IO upwelling.

Supplemental information related to this paper is available at the Journals Online website: https://doi.org/10.1175/JCLI-D-19-0386.s1.

© 2020 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Xiaolin Zhang, xiaolin.zhang@colorado.edu

Abstract

This paper investigates interannual variability of the tropical Indian Ocean (IO) upwelling through analyzing satellite and in situ observations from 1993 to 2016 using the conventional Static Linear Regression Model (SLM) and Bayesian Dynamical Linear Model (DLM), and performing experiments using a linear ocean model. The analysis also extends back to 1979, using ocean–atmosphere reanalysis datasets. Strong interannual variability is observed over the mean upwelling zone of the Seychelles–Chagos thermocline ridge (SCTR) and in the seasonal upwelling area of the eastern tropical IO (EIO), with enhanced EIO upwelling accompanying weakened SCTR upwelling. Surface winds associated with El Niño–Southern Oscillation (ENSO) and the IO dipole (IOD) are the major drivers of upwelling variability. ENSO is more important than the IOD over the SCTR region, but they play comparable roles in the EIO. Upwelling anomalies generally intensify when positive IODs co-occur with El Niño events. For the 1979–2016 period, eastern Pacific (EP) El Niños overall have stronger impacts than central Pacific (CP) and the 2015/16 hybrid El Niño events, because EP El Niños are associated with stronger convection and surface wind anomalies over the IO; however, this relationship might change for a different interdecadal period. Rossby wave propagation has a strong impact on upwelling in the western basin, which causes errors in the SLM and DLM because neither can properly capture wave propagation. Remote forcing by equatorial winds is crucial for the EIO upwelling. While the first two baroclinic modes capture over 80%–90% of the upwelling variability, intermediate modes (3–8) are needed to fully represent IO upwelling.

Supplemental information related to this paper is available at the Journals Online website: https://doi.org/10.1175/JCLI-D-19-0386.s1.

© 2020 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Xiaolin Zhang, xiaolin.zhang@colorado.edu

1. Introduction

In the tropical Pacific and Atlantic Oceans, easterly trade winds prevail and upwelling occurs in the eastern equatorial basin. In contrast, the Indian Ocean is subject to intense seasonally reversing monsoon wind forcing, and the annual mean prevailing winds in the equatorial basin are westerlies (Figs. 1a–c). The negative wind stress curl (positive Ekman pumping velocity) associated with the southeasterly trades south of 10°S and westerlies near the equator induces open-ocean upwelling in the western basin around 12°–2°S, where thermocline shoals as a ridge (McCreary et al. 1993; Murtugudde et al. 1999), as shown by the shallow depth of 20°C isotherm (D20; Fig. 1a, the boxed region in the west). This region is referred to as the Seychelles–Chagos thermocline ridge (SCTR; Hermes and Reason 2008; Yokoi et al. 2008, 2009), which is the mean upwelling zone of the tropical Indian Ocean. In this paper, we refer it to as the west Indian Ocean (WIO) upwelling hereafter. During boreal summer, coastal upwelling occurs along the coasts of Sumatra and Java in the eastern equatorial basin and near the coasts of Somalia, Oman, and southwest India in the north Indian Ocean (Fig. 1c; Susanto et al. 2001; Schott et al. 2009). During the years of positive Indian Ocean dipole (IOD; Saji et al. 1999; Webster et al. 1999), intense upwelling occurs over the southeast tropical Indian Ocean in boreal fall, which significantly amplifies the seasonal upwelling along the coasts of Java and Sumatra (see Figs. 1d,e). This upwelling resides in the tropical east Indian Ocean (EIO) warm pool region, and hereafter is referred to as EIO upwelling.

Fig. 1.
Fig. 1.

(a) Annual mean, (b) January, and (c) September climatology of depth of 20°C isotherm (D20; color) and ERA-Interim surface wind stress (arrows) for the 1993–2016 period. The boxed region in the west shows the Seychelles–Chagos thermocline ridge (SCTR) region (2°–15°S, 50°–80°E), the mean upwelling zone of the tropical Indian Ocean; the boxed area in the east shows the eastern node of the Indian Ocean dipole (IOD) area (10°S–0°N, 90°–115°E), which includes the seasonal upwelling region along the coasts of Sumatra and Java during boreal summer–fall. (d) September–November (SON) averaged D20 anomaly (D20A) and wind stress anomaly for 1997, when positive IOD (pIOD) and El Niño co-occurred; (e) as in (d), but for 2012, an independent pIOD (no El Niño or La Niña); (f) as in (d), but for 2009, an El Niño year (no IOD). The D20 is based on the monthly ECMWF Ocean Reanalysis System 4 (ORAS4) data from 1993 to 2016.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

The WIO upwelling exhibits significant interannual variability, with large variability occurring in boreal winter (Xie et al. 2002; Huang and Kinter 2002). Existing studies suggest that interannual D20 anomaly (D20A) in the WIO upwelling zone is primarily caused by Rossby waves driven by winds in the central and eastern Indian Ocean (e.g., Tozuka et al. 2010; Trenary and Han 2012), but local Ekman pumping is not negligible (Tozuka et al. 2010). The influence of remote forcing from the Pacific via the Indonesian Throughflow is weak in this area, even though it has significant contributions in the southeast Indian Ocean (e.g., Potemra 2001; Trenary and Han 2012; Deepa et al. 2018; Hu et al. 2019). Indeed, the importance of Rossby waves in causing interannual variability of sea level over the south Indian Ocean has been extensively studied (e.g., Woodberry et al. 1989; Périgaud and Delecluse 1992, 1993; Fu and Smith 1996; Masumoto and Meyers 1998; Yang et al. 1998; Chambers et al. 1999; White 2001; Birol and Morrow 2001; Wang et al. 2001; Jury and Huang 2004; Baquero-Bernal and Latif 2005; Rao and Behera 2005; Zhuang et al. 2013).

Since El Niño–Southern Oscillation (ENSO) and the IOD are the most influential climate modes over the tropical Indian Ocean, efforts have been made to understand their influences on interannual variability of the Indian Ocean. Correlation, regression, and composite analyses show that winds over the Indian Ocean that drive the Rossby waves are associated with the ENSO (e.g., Huang and Kinter 2002; Xie et al. 2002), and ENSO dominates the wind-driven Rossby waves south of ~10°S whereas the IOD dominates north of 10°S (e.g., Rao and Behera 2005; Yu et al. 2005; Gnanaseelan and Vaid 2010). Composites for pure El Niño, pure positive IOD (pIOD), and their co-occurrence years, however, show larger sea level variability during pure pIOD compared to pure El Niño years both north and south of 10°S, and their co-occurrence significantly enhances the variability magnitudes (Deepa et al. 2018). In terms of upwelling in the EIO, ocean model experiments show that interannual anomaly of upwelling is forced by winds both remotely from the equatorial Indian Ocean and locally within the EIO for the 1997 IOD–El Niño co-occurrence (Murtugudde et al. 2000; Han and Webster 2002) and for the entire 2001–11 period (Chen et al. 2016). Correlation analyses suggest that ENSO is the dominant cause for interannual variability of coastal upwelling along the Sumatra and Java coasts (e.g., Susanto et al. 2001), whereas others suggest that the IOD is more important than ENSO in causing the EIO upwelling (e.g., Shinoda et al. 2004; Yu et al. 2005; Chen et al. 2016).

The inconsistent results from existing studies suggest that the relative importance of ENSO and IOD in causing interannual variability of upwelling in different regions of the Indian Ocean is inconclusive. While empirical analyses (e.g., correlation) from previous studies qualitatively suggested the importance of winds associated with ENSO and IOD in driving interannual variability over the Indian Ocean, the relative importance of ENSO versus IOD in causing the WIO and EIO upwelling has not been quantified. Even for ENSO alone, the space–time complexity exists in terms of magnitude, spatial pattern, and temporal evolution (Timmermann et al. 2018). There are two distinct types of ENSO events: the central Pacific ENSO (CP ENSO) with a center—the location of the maximum sea surface temperature anomaly (SSTA)—located in the central equatorial Pacific (5°S–5°N, 160°E–150°W), and the eastern Pacific ENSO (EP ENSO) with a center located in the eastern equatorial Pacific (5°S–5°N, 150°–90°W) (Takahashi et al. 2011; Cai et al. 2018). Even though the convection maxima associated with the CP and EP ENSO events are located in different regions, it is unclear whether or not they have different impacts on the Indian Ocean.

The goal of this paper is to utilize satellite altimeter data, in situ observations combined with ocean model experiments, and application of advanced statistic tool, the Bayesian dynamical linear model (DLM), to provide a thorough investigation on the roles played by ENSO and the IOD, together with CP versus EP ENSO, in causing interannual variability of the WIO and EIO upwelling. The Bayesian DLM is able to capture the time-varying impacts of ENSO and IOD on Indian Ocean upwelling, which cannot be captured by the static linear regression model (section 2d). The associated wind-driven wave dynamics are examined using a linear continuously stratified ocean model (LOM). Our main period of interest is from 1993 to 2016, when the satellite altimeter data are available. Our analyses also go back to 1979, when satellite observed outgoing longwave radiation (OLR) and ocean–atmosphere reanalysis data are available.

Since variability of upwelling directly affects biogeochemical processes and is tied to variability of the thermocline, SST, and upper ocean heat content, results from this study have important implications for understanding climate variability and marine ecosystems. The structure for the rest of the paper is as follows. Section 2 describes the data and approach. Section 3 reports our research results, and section 4 provides a summary and discussion.

2. Data and approach

a. Datasets for detecting upwelling

Based on the definition of upwelling, wind-driven surface Ekman divergence causes thermocline to shoal, bringing the colder subsurface water to the surface layer. Consequently, the depth of the thermocline, which is often represented by the depth of 20°C isotherm, is used for detecting upwelling. In this paper, D20 is calculated from the 1° × 1° monthly temperature data of ECMWF Ocean Reanalysis System 4 (ORAS4) available for 1958–2016 (Balmaseda et al. 2013). The 1° × 1° gridded temperature data from the gridpoint value (GPV) of the monthly objective analysis using the Argo data (MOAA; Hosoda et al. 2008) are also used to calculate D20 for its available period of 2001 to the present and compared with that of ORAS4 data. Since the tropical ocean is dominated by baroclinic response (e.g., Fukumori et al. 1998), a shallower (deeper) thermocline corresponds to a lower (higher) sea surface height (SSH), and therefore SSH anomaly (SSHA) can also serve as an indicator for upwelling. The SSHA datasets are the 1/4° monthly Archiving, Validation, and Interpretation of Satellite Oceanography (AVISO) data (https://www.aviso.altimetry.fr/en/data.html) from 1993 to 2016 and the 1° × 1° monthly ORAS4 reanalysis, which assimilated altimetry data. The monthly upper 700-m thermosteric sea level data from World Ocean Atlas 2013 (WOA13; Levitus et al. 2012) from 1993–2015 are also analyzed. The interannual D20A and SSHA are calculated by removing their monthly climatologies and linear trends for the periods of interested.

To understand the processes associated with the climate modes’ impacts on interannual variability of WIO and EIO upwelling, we also analyze the 0.25° × 0.25° monthly cross-calibrated multiplatform (CCMP) satellite ocean surface wind vectors available for July 1987–June 2011 (Atlas et al. 2011); the 0.75° × 0.75° monthly ERA-Interim winds available from 1979 to 2016 (Dee et al. 2011); the 2.5° × 2.5° monthly CMAP precipitation available for 1979–2017 (Xie and Arkin 1996, 1997), and 2.5° × 2.5° NOAA outgoing longwave radiation (OLR) for 1979–2017 (Chelliah and Arkin 1992). The 1° × 1° HadISST sea surface temperature (SST) data (Rayner et al. 2003) available since 1870 are used to calculate the climate modes’ indices and to reveal interannual variability of upwelling cooling.

b. Climate modes

Since the ENSO and IOD are the two major climate modes that induce interannual variability over the Indian Ocean, in this study we focus on quantifying their contributions. ENSO is represented by the Niño-3.4 index, which is the monthly SST anomaly (SSTA) averaged in the 5°N–5°S, 170°–120°W region. The IOD is documented by dipole mode index (DMI), which is defined as the monthly SSTA difference between the western pole (10°S–10°N, 50°–70°E) and eastern pole (10°S–0°, 90°–110°E), following Saji et al. (1999).

Figure 2a shows the time series of monthly Niño-3.4 index (red curve) and DMI (blue curve) from January 1993 to December 2016, which are used for assessing the ENSO and IOD effects on Indian Ocean upwelling (sections 2ce below). The El Niño and positive IOD events that occurred during the 1979–2016 period are listed in Table 1, based on Yu et al. (2012) and the NOAA website (https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php) for CP and EP El Niño events and the Australian government Bureau of Meteorology website (http://www.bom.gov.au/climate/iod/) for pIOD events.

Fig. 2.
Fig. 2.

(a) Monthly Niño-3.4 index (red line) and DMI (blue line) calculated from monthly HadISST with their linear trends and monthly climatology of 1993–2016 removed. Units: °C. Their correlation coefficient and standard deviations (STD) for each time series are shown at the bottom of the panel. (b) Time series of monthly DMI (red line), same as the blue curve in (a), and independent DMI with ENSO effect removed using conventional static linear regression (blue line). In each panel, the dash–dotted horizontal red (blue) lines show ±1 STD of the corresponding index. All correlation coefficients exceed 95% significance. Note that here and elsewhere in this paper, significance estimates for correlation coefficients are based on Ebisuzaki (1997).

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

Table 1.

Major El Niño events (second column) and positive IOD events (third column) as identified based on the NOAA website (https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php) and the Australian government Bureau of Meteorology website (http://www.bom.gov.au/climate/iod/) for the period of January 1979 to December 2016. The types of the El Niño events are identified by the majority consensus method as described in Yu et al. (2012). Boldface indicates that the EP and pIOD events co-occur and the italic font means that the CP and pIOD events co-occur.

Table 1.

c. The static linear model

In a conventional static multiple linear regression model, a response variable Y is equated to a linear function of independent predictors (X1, X2, …, XN), that is,
Y(t)=b0+b1X1(t)+b2X2(t)++bNXN(t)+εt,
where each coefficient bi (i = 0, 1, 2, …, N) is a constant and represents the influence of a unit change in Xi (i = 1, 2, …, N) on Y, and εt represents an error term. In this paper, we use the ENSO and IOD index respectively as a single predictor. Equation (1) becomes
Y(t)=b0+b1X1+εt,
where X1 represents the Niño-3.4 index or DMI, and Y(t) represents time series of upwelling indicator (e.g., D20A, SSHA, SSTA) at a specific location or averaged over a region (e.g., the EIO or WIO). Note that we do not perform regressions onto the ENSO and IOD indices simultaneously because they are correlated (r = 0.43; >95% significance; Fig. 2a) and thus are not independent. Therefore, our results reflect the maximum amount of variance that might be attributed to ENSO and IOD. Here, we also calculated an ENSO-independent DMI, referred to as partial-DMI (pDMI), by removing the regression of DMI onto Niño-3.4 index from the original DMI. Note that pDMI and DMI are highly correlated (r = 0.90, >95% significance), and the ratio of their standard deviations (STD), std(pDMI)/std(DMI), is 0.90. The major differences between pDMI and DMI (blue and red lines of Fig. 2b) are seen for ENSO and IOD co-occurring years (e.g., 1997, 2015, and 2006). Han et al. (2017a, 2018) removed the decadal ENSO effect from the decadal DMI to make the two indices independent, and then performed regressions onto the ENSO and IOD indices simultaneously. The underlying assumption is that ENSO has an active impact on IOD, but the IOD does not affect ENSO. Existing studies however show that IOD also has an active influence on ENSO (e.g., Izumo et al. 2010). Therefore, removing the ENSO effect from the DMI prior to the regression analysis may overestimate (underestimate) the ENSO (IOD) effect, as discussed in Han et al. (2017a, 2018). Due to this limitation, we chose to use a single predictor in this paper [i.e., Eq. (2)].

d. The Bayesian dynamic linear model

Since the regression coefficients in the static linear model (SLM) are constant and do not vary with time within the temporal period examined, SLM measures the stationary influence of the predictor on the predictand. In real climate system, however, the relationship between the predictor and predictand is often nonstationary. For instance, it has been shown that the influence of El Niño on the subtropical northwest Pacific climate has significantly increased after the 1970s (Xie et al. 2010), ENSO–IOD co-occurrence has an apparent increase after 1976 (Annamalai et al. 2005), the ENSO–Indian summer monsoon relation has broken down during recent decades (Kumar et al. 1999; Ashok et al. 2004), and the IOD effect on Indian summer monsoon has strengthened in recent decades compared to ENSO (Krishnaswamy et al. 2015). Recent studies have shown that decadal variability of ENSO, IOD, and Indian summer monsoon have varying impacts on decadal variability of Indian Ocean Walker cell and sea level (Han et al. 2017a,b, 2018).

The Bayesian DLM allows coefficients bi to vary with time, which overcomes the limitation of constant coefficients of the SLM and thus can measure dynamical (time evolving) impacts of Xi on Y. Below we use a single predictor X1, which represents either Niño-3.4 index or DMI, as an example. The DLM consists of two equations: an “observation equation” analogous to the SLM, as shown by Eq. (3) below, and a “state equation” that controls the dynamic evolution of coefficients bi, represented by Eq. (4):
Y(t)=b0(t)+b1(t)X1(t)+ε(t),whereε(t)~N[0,V(t)],
bi(t)=bi(t1)+wi(t),wherewi(t)~N[0,Wi(t)],i=0,1.

The state Eq. (4) means that the predictive distribution of bi at each time step t (i.e., posterior) is updated based on its previous step t − 1 distribution (i.e., prior) and the probability of observations Y conditional on bi at time t (i.e., the likelihood) using Bayes’ theorem (Petris et al. 2009). Coefficient bi (i = 0, 1) is obtained by applying Kalman filtering and smoothing, with the corresponding SLM coefficient as its initial guess. In Eqs. (3) and (4), the b0(t) term represents a time-varying level or intercept whose variability is unexplained by predictor X1, while the b1 term represents the nonstationary influence of X1 on Y; ε(t) and wi(t) are independent white noises or errors, distributed normally with a mean of 0 and variances of V(t) and Wi(t). For more details on the Bayesian DLM and its Indian Ocean applications, please see Petris et al. (2009), Petris (2010), R Development Core Team (2016), and Han et al. (2017a).

e. The linear ocean model and experiments

Since both the SLM and Bayesian DLM examine the simultaneous relationship between the predictor (ENSO or IOD) and predictand (e.g., D20A, SSHA), oceanic wave propagations in the ocean are not properly captured. To provide insight into the effect of wind-driven oceanic Kelvin and Rossby waves, we also performed model experiments using the linear continuously stratified ocean model (LOM; McCreary 1981; McCreary et al. 1996; Han et al. 2004, 2011; Han 2005). The equations of motion are linearized about a background state of rest with stratification represented by Brunt–Väisälä frequency derived from observations, and the ocean bottom is assumed flat at 4000 m (see the appendix for more details about the model). Solutions are found for each baroclinic mode, and the total solution is represented by the sum of the first 25 baroclinic modes, which are well converged with this choice.

The LOM is configured for the tropical Indian Ocean north of 29°S with a horizontal resolution of 55 km × 55 km. It was first spun up for 20 years using monthly climatology of Cross-Calibrated Multi-Platform (CCMP) version 2.0 (CCMP2) wind stress (Wentz et al. 2015) forcing, and then integrated forward in time forced by the monthly CCMP2 wind stress from 1988 to 2016. This solution is referred to as LOM main run (LOM_MR). To isolate the effects of ENSO and the IOD, respectively, we first extracted the Indian Ocean surface wind stress signals associated with ENSO and IOD. Specifically, we applied the Bayesian DLM with the monthly Niño-3.4 index and DMI respectively as the predictor, and monthly CCMP2 wind stress at each grid point of the Indian Ocean as the predictand. Since it only takes a few days for the Indian Ocean surface wind to respond to the SSTA associated with ENSO and the IOD, the simultaneously relationships between the monthly climate indices and surface wind anomalies in the Bayesian DLM are valid. Then we used the ENSO and IOD wind stress to force the LOM, and obtained experiments LOM_ENSO (forced by ENSO-related wind stress) and LOM_IOD (forced by IOD-related wind stress). These solutions quantitatively estimate the ENSO and IOD impacts, including the wind-driven oceanic wave processes that are not properly represented by the statistical analyses using SLM and Bayesian DLM.

To assess the effects of direct wind forcing over the ocean interior versus the effect of the eastern boundary (i.e., coastal Kelvin waves and reflected Rossby waves), we carried out an additional experiment in which a “damper” was added to the LOM, denoted by LOM_DAMP; otherwise it is the same as the LOM_MR. The damper efficiently absorbs the energy of the incoming equatorial Kelvin waves, and thus no coastal Kelvin waves are generated along the coasts and no Rossby waves are reflected back into the ocean interior from the eastern boundary (McCreary et al. 1996). The solution difference (LOM_MR − LOM_DAMP) isolates the effects of coastal Kelvin waves plus reflected Rossby wave effects.

3. Results

In this section, we first document the interannual variability of upwelling over the Indian Ocean (section 3a). Then we examine the effects of ENSO and IOD on the Indian Ocean upwelling, particularly in the SCTR mean upwelling zone and EIO boreal summer seasonal upwelling region using SLM (section 3b) and Bayesian DLM (section 3c). Finally, we assess the wind-driven oceanic waves (e.g., Rossby waves) in affecting the Indian Ocean upwelling by conducting model experiments using the LOM (section 3d).

a. Observed upwelling

To reveal the spatial pattern of the interannual variability of upwelling in the tropical Indian Ocean, we calculated the STD of monthly D20A and SSHA based on observational datasets and reanalysis products (Fig. 3). The ORAS4 data show large-amplitude interannual variability of D20A in the mean upwelling zone of the SCTR region for both the 1993–2016 and 2001–16 periods, with STD exceeding 15 m in the WIO area (2°–15°S, 50°–80°E) and maximum of ~25 m in the eastern part of the SCTR (5°–16°S, 66° to 88°E), referred to as the Max Region hereafter (western and central boxes of Figs. 3a,b; Figs. 1a–c). Large variability (10–15 m) also occurs in the eastern equatorial basin, including the EIO box region (0°–10°N, 90°–115°E). MOAA D20A from 2001–16 shows similar variability pattern (Fig. 3c), but the magnitudes are evidently smaller than those of ORAS4, likely due to the lack of spatial coverage of Argo floats in the MOAA dataset.

Fig. 3.
Fig. 3.

Standard deviation (STD) of monthly anomalies of D20 (units: m) based on (a) ORAS4 reanalysis from 1993–2016, (b) ORAS4 from 2001–16, and (c) MOAA-GPV from 2001–16. Also shown are the STD maps of monthly (d) WOA13 upper 700-m thermosteric sea level anomaly (SLA; cm) for 1993–2015, (e) ORAS4 sea surface height anomalies (SSHA; cm) for 1993–2016, and (f) AVISO SSHA for 1993–2016. The black dashed boxes show the east Indian Ocean (EIO) and west Indian Ocean (WIO) mean upwelling zones, the same boxes as in Fig. 1. The black solid boxes show the maximum STD region based on both ORAS4 and AVISO SSHA (5°–16°S, 66° to 88°E).

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

Since interannual variability of SSHA over the tropical oceans is dominated by baroclinic response (e.g., Fukumori et al. 1998), SSHA mirrors D20A (e.g., Gill 1982). Therefore, SSHA can also serve as an excellent indicator for tropical Indian Ocean upwelling. The STD maps of ORAS4 and AVISO satellite data exhibit similar amplitudes and spatial patterns (Figs. 3e,f), with strong variability occurring in the SCTR region and EIO, the same areas as those of D20A. Exceptions are for the western Bay of Bengal, the western Arabian Sea and the southwestern Indian Ocean Agulhas Current region where large SSHAs appear in satellite data but not in ORAS4 reanalysis product. This is because strong oceanic instabilities occur in these areas (e.g., Li and Han 2015), which are detected by the 1/4° × 1/4° AVISO data but are missed by the 1° × 1° ORAS4 reanalysis. Note that ORAS4 assimilated satellite SSH data (Balmaseda et al. 2013), which contributes to their good agreement. The upper 700-m thermosteric sea level anomalies of WOA13 were also analyzed (Fig. 3d), but their variability magnitudes are much smaller than that of ORAS4 and AVISO satellite data, suggesting that deep ocean variability below 700 m and halosteric sea level are also important in causing interannual sea level variability.

To understand the relationship between the upwelling in the WIO and EIO regions, we first calculated the D20A and SSHA averaged in the WIO and EIO (Fig. 4). Evidently, upwelling anomalies in the WIO and EIO are out of phase, with simultaneous correlation of D20A being −0.66 (>95% significance) and comparable STD values. This out-of-phase relationship is also captured by SSHA, showing correlation of r = −0.68 (>95% significance) for ORAS4 reanalysis and r = −0.65 for AVISO observation (>95% significance), with similar variability amplitudes (STD values) in the WIO and EIO. Since SSTA can be affected by upwelling cooling, it may also be used as a proxy of upwelling. Different from the D20A and SSHA, however, the monthly SSTA averaged in the WIO and EIO are positively correlated, with r = 0.32 (>95% significance) from 1993 to 2016. This is because the monthly SSTA is overall dominated by the Indian Ocean basinwide warming/cooling induced by ENSO via surface heat fluxes (Klein et al. 1999). In contrast, the SSTA averaged for SON (dashed lines in Fig. 4d), the peak period of the IOD, shows a WIO–EIO correlation of r = −0.62 (>95% significance), suggesting that interannual SSTA can serve as an upwelling indicator in the WIO and EIO only during the boreal fall season.

Fig. 4.
Fig. 4.

Time series of monthly (a) ORAS4 D20A, (b) ORAS4 SSHA, (c) AVISO SSHA, and (d) SSTA from HadISST averaged in the WIO (red line) and EIO (blue line) mean and seasonal upwelling zones (boxed regions of Fig. 1). In (d), the two dotted lines show the corresponding SON mean SSTA. The correlation and STD values are shown at the bottom of each panel. (e) Time series of D20A averaged over the maximum D20A region (central box of Fig. 3); (f) as in (e), but for ORAS4 and AVISO SSHA.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

The lead/lag correlations between the time series of the WIO and EIO show a maximum correlation of −0.73 for AVISO SSHA, when WIO lags by 2 months. The lag of upwelling in the west indicates that signals generated in the EIO propagate westward via Rossby waves, as pointed out by previous studies (section 1) and further demonstrated below. In the maximum D20A and SSHA region of the central tropical south Indian Ocean (i.e., the Max Region; solid box of Fig. 3), the averaged time series shows a STD of 9.96 m for ORAS4 D20A (Fig. 4e) and a STD of 5 cm (5 cm) for ORAS4 (AVISO) SSHA. These values are larger than those of the WIO and EIO (Fig. 4). Since SSHA can well capture the upwelling features, hereafter we use the AVISO SSHA as upwelling proxy.

To examine the seasonality of the interannual variability of upwelling, Fig. 5a (left column of Fig. 5) shows the bimonthly STD maps of AVISO SSHA for the 1993–2016 period. In the EIO, interannual SSHAs near the coasts of Sumatra and Java are phase-locked with the seasonal upwelling, being strong from July to November (Fig. 5a) and beginning to decay in December (not shown). Accompanying the strong SSHA variability near the Sumatra and Java coasts there are large-magnitude SSHAs appearing in June over the eastern Indian Ocean near 10°S, 100°E, which is to the southwest of the EIO upwelling region. It then strengthens and progresses westward from July to March. The SSHA amplitudes enhance within the WIO area from January to March, suggesting that local forcing over the SCTR region strengthens the upwelling variability in this season, in addition to the remote influence from Rossby waves generated in the eastern basin, consistent with Tozuka et al. (2010).

Fig. 5.
Fig. 5.

Bimonthly STD maps of AVISO SSHA from 1993–2016 for (a) total SSHA, (b) SSHA due to ENSO effect (b1 × ENSO), and (c) SSHA due to IOD effect (b1 × IOD) from the Bayesian dynamical linear model (DLM). The boxes are the same as in Fig. 3.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

b. ENSO and IOD impacts on upwelling: Results from the conventional SLM

To understand the roles of ENSO and IOD in causing interannual variability of the Indian Ocean upwelling, we begin with examining their effects using the conventional SLM [i.e., Eq. (2)]. For the EIO, ENSO has a significant contribution to the satellite-observed upwelling represented by SSHA (black and blue lines of Fig. 6a), with a correlation of r = 0.56 (>95% significance) and STD ratio of STD(slm)/STD(obs) = 0.56 from 1993 to 2016. The correlation between the IOD-contributed and observed upwelling is r = 0.68 and the STD ratio is STD(slm)/STD(obs) = 0.67 (Fig. 6d). These results suggest that for the upwelling in the EIO, the eastern node of the IOD that resides in the warm pool, both ENSO and IOD are important, with IOD playing a somewhat larger role based on the SLM. For the WIO mean upwelling region, both ENSO and IOD are important, but ENSO plays a larger role. The correlation for the ENSO-contributed and satellite-observed SSHA is r = 0.62 and the STD ratio is STD(slm)/STD(obs) = 0.61, compared to the r = 0.46 and STD(slm)/STD(obs) = 0.45 of the IOD contribution (Figs. 6b,e). Similar calculations were done for the Max Region (solid boxes of Figs. 3 and 5), and the results show that ENSO plays a more important role than the IOD, with r = 0.65 (0.41) and STD ratio = 0.64 (0.42) for ENSO (IOD).

Fig. 6.
Fig. 6.

Time series of satellite-observed AVISO SSHA (black line), SSHA explained by ENSO using the conventional static linear model (SLM; blue line) and Bayesian DLM (red line), for the (a) EIO box, (b) WIO box, and (c) maximum SSHA box shown in Figs. 3 and 5. (d)–(f) As in (a)–(c), but for SSHA explained by the IOD. Specifically, the blue and red lines are the b1X1 terms of Eqs. (2) and (3), respectively, with X1 being the Niño-3.4 index for (a)–(c) and DMI for (d)–(f). The STD for each time series and the correlations between AVISO observed SSHA and ENSO- or IOD-explained SSHA using SLM (Bayesian DLM) are shown at the bottom of each panel. All correlation coefficients exceed 95% significance.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

c. ENSO and IOD impacts on upwelling: Results from the Bayesian DLM

1) ENSO and IOD impacts and causes

Compared to the SLM, the Bayesian DLM is able to capture the time-varying influences of ENSO and IOD on interannual variability of tropical Indian Ocean upwelling (section 2d). Indeed, the correlation coefficients increased in the Bayesian DLM (red curves of Figs. 6a–c), with ENSO-contributed and satellite observed SSHA correlations being 0.70, 0.79, and 0.83 in the EIO, WIO, and Max Region, respectively, compared to the 0.56, 0.62, and 0.65 from the SLM. Similarly, the IOD-contributed and observed SSHA correlations are 0.79, 0.59, and 0.58 in the three regions compared to 0.68, 0.46, and 0.41 in the SLM (Figs. 6d–f). For the SSHA variance, the most notable change from the SLM is that ENSO explains significantly larger upwelling variance in the EIO, with the STD(dlm)/STD(obs) ratio being 0.79 (0.72) for ENSO (IOD) in the Bayesian DLM, compared to 0.56 (0.67) in the SLM. Considering the correlations of 0.70 (0.79) between observed SSHA and ENSO-contributed (IOD-contributed) SSHA in the EIO region, we conclude that ENSO and the IOD play a comparable role in determining the interannual variability of the EIO upwelling for the 1993–2016 period. The overall conclusions on the larger contributions of ENSO than IOD in the WIO and Max Region, however, remain unchanged.

To reveal the overall spatial patterns of SSHA caused by ENSO and IOD, respectively, Fig. 7 shows the leading empirical orthogonal function (EOF) mode (EOF1) of atmospheric and oceanic variability caused by ENSO (Figs. 7a–c) and IOD (Figs. 7d–f) from 1993 to 2016 using the Bayesian DLM. The EOF1 captured the majority of the variance for each field, explaining 64.3% (52.4%) for SSHA, 74.8% (74.0%) for anomalous Ekman pumping velocity, 49.6% (56.6%) for precipitation anomaly, and 64.0% (61.5%) for outgoing longwave radiation anomaly (OLRA) associated with ENSO (IOD). Note that we obtained the EOFs for each field separately; however, their PCs are highly correlated. For instance, for the ENSO-related fields, the correlation between the PC1s is 0.97 for SSHA and precipitation, 0.98 for SSHA and OLRA, 0.98 (0.92) for SSHA and τx(τy), and 0.98 for SSHA and Ekman pumping velocity. For the IOD-related fields, the correlation between the PC1s is 0.93 for SSHA and precipitation, 0.95 for SSHA and OLRA, 0.97 (0.92) for SSHA and τx(τy), and 0.96 for SSHA and Ekman pumping velocity. All of the correlations exceeded 95% significance.

Fig. 7.
Fig. 7.

The leading EOF mode (EOF1) of atmospheric and oceanic variability caused by (a)–(c) ENSO and (d)–(f) the IOD from 1993 to 2016 using Bayesian DLM. Color shading shows EOF1 of (a) SSHA, (b) precipitation anomaly, and (c) outgoing longwave radiation anomaly (OLRA) induced by ENSO. Values within parentheses show variance explained by EOF1. The EOF1s of surface wind stress (vectors) and Ekman pumping velocity (we, line contours), defined as we = Curl(τ/ρwf) = (∂/∂x)(τy/ρwf) − (∂/∂y)(τx/ρwf), are also plotted in each panel. (d)–(f) As in (a)–(c), respectively, but for IOD effect. Unit for we is 1 × 107 m s−1, and positive (negative) values are in red (green), with values of 3, 6, −24, −18, −12, and −6. The units for SSHA, precipitation anomaly, OLRA, and anomalous wind stress are cm, mm day−1, W m−2, and N m−2, respectively.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

Evidently, ENSO and IOD induce similar upwelling (i.e., SSHA) patterns overall, with El Niño and pIOD causing upwelling in the eastern basin and downwelling in the central and western basins of the tropical south Indian Ocean (Figs. 7a,d). This is because El Niño and pIOD are associated with surface southeasterly and equatorial easterly winds in the EIO, which drive coastal and equatorial upwelling, shoal the thermocline and lower sea level, producing negative SSHA and SSTA in the EIO. The Bjerknes feedbacks, together with the reduced convection and precipitation in the EIO warm pool but enhanced convection and precipitation in the western tropical basin (Figs. 7b,c,e,f), enhance the easterlies and thus intensify the EIO upwelling. On the other hand, the similar ENSO- and IOD-related upwelling patterns may be partly due to their co-occurrence. Among the seven El Niño and five pIOD events that occurred during 1993–2016, four of them co-occurred (Table 1 and Fig. 8); however, the surface wind and SSHA patterns for individual pure pIOD and pure El Niño years show similar patterns (Fig. S1 in the online supplemental material), and their composites during 1979–2016 are also similar, resembling those of the El Niño–pIOD co-occurrence years (Fig. S2). These results suggest that the dynamical reason discussed above is the major cause for their similarities.

Fig. 8.
Fig. 8.

Time series of Bayesian DLM coefficients, that is, b1 of Eq. (3), in the EIO (eastern box of Fig. 5) for (a) ENSO (red curve), and (b) the IOD (blue curve). (c),(d) As in (a),(b), but for the WIO upwelling region (western box of Fig. 5). (e),(f) As in (a),(b), but for the central box of Fig. 5 where SSHA reaches the maximum. The horizontal dashed–dotted black line in each panel shows the corresponding SLM coefficient, that is, b1 of Eq. (2), which is a constant value. The El Niño (La Niña) years are marked as yellow dots (green diamonds) in (a), (c), and (e), and “×” is added to the yellow dots for EP El Niños. Positive (negative) IOD events are marked as red dots (green diamonds) in (b), (d), and (f).

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

Upwelling is the strongest near the eastern boundary, due to the eastward propagation of equatorial Kelvin waves (e.g., Chen et al. 2016); upon impinging on the eastern boundary, the upwelling signals (e.g., SSHAs) propagate northward into the Bay of Bengal and southward to the coasts of Sumatra and Java (Figs. 7a,d), enhancing the seasonal coastal upwelling there (Fig. 1c).

Associated with the equatorial easterlies are the off-equatorial negative Ekman pumping velocities (we, green and red curves of Figs. 7a,d) associated with both ENSO and IOD, which deepen the thermocline and raise sea level. The positive SSHAs propagate westward as Rossby waves (e.g., Webster et al. 1999; Xie et al. 2002; Fig. 5a), reducing the mean upwelling in the SCTR region (Fig. 1). Note that in the ocean interior, ENSO causes largest SSHA in the Max Region, with El Niño inducing positive SSHA there (Fig. 7a). This is because SSHA, which mirrors D20A (Fig. 4), is roughly proportional to the magnitude of integrated we along Rossby wave’s characteristics (from east to west), attaining its maximum where we changes sign from negative to positive (green curves of Fig. 7a). Similarly, the IOD also causes the maximum SSHA in the Max Region, but the SSHA amplitude is much weaker than that of ENSO, with ENSO (IOD) associated SSHA STD being 3.3 cm (1.9 cm) in the Max Region compared to 2.8 cm (1.8 cm) in the WIO (Fig. 6). The stronger influence of ENSO over the WIO and Max Region, particularly during October–December (Fig. 5), results from the stronger ENSO-related negative Ekman pumping velocity compared to that of the IOD (not shown). Because the simultaneous relationships between monthly SSHA and climate indices in the SLM and Bayesian DLM may not fully capture Rossby wave propagation (section 2e), the propagation effect of Rossby waves will be further discussed in section 3d.

The impacts of ENSO and IOD have apparent seasonality. During boreal winter and spring (January and March panels of Figs. 5b,c), ENSO dominates the IOD in causing anomalous SSHAs and upwelling. Therefore, the enhanced SSHAs in the SCTR mean upwelling zone (both the WIO and Max Region) observed during boreal winter and spring result mainly from ENSO effects (cf. Figs. 5a–c). From May to November, ENSO and the IOD have comparable influences.

2) The time-varying impacts of ENSO and the IOD

Since the coefficients of Bayesian DLM vary with time (Fig. 8), they reflect the time-evolving impacts of ENSO and IOD on upwelling. Larger b1 magnitude indicates that for the same ENSO strength, the ENSO impact on the Indian Ocean upwelling is stronger. During the four El Niño–pIOD co-occurring years of 1994, 1997, 2006, and 2015 (Table 1; Fig. 8), El Niño and pIOD events are not independent, and thus stronger (weaker) El Niño effects generally correspond to stronger (weaker) pIOD impacts. For instance, larger (smaller) DLM coefficients are found for both ENSO and IOD in 1997 and 1994 (2015) (Figs. 8a,b). For 2006, however, the large DLM coefficient for ENSO (larger than that of 1994) corresponds a moderate DLM coefficient for IOD (smaller than that of 1994), indicating that the IOD has its independent aspect even though it co-occurred with ENSO. Interestingly, while the Niño-3.4 index has similar magnitude for 1997 and 2015 (Fig. 2a), the ENSO coefficient b1 for 1997 is apparently larger than that of 2015 particularly in the EIO (Figs. 8a,c,e), and the SSHAs associated with 1997 El Niño are larger than those of 2015 (Figs. 6a–c).

Why does the 1997 El Niño have a much stronger impact than the 2015 El Niño? To answer this question, we analyzed the SSTA, OLRA, surface wind stress, and Ekman pumping velocity anomalies for 1997 and 2015 (Fig. 9). Compared to 1997, warmer SSTA occurred in the central equatorial Pacific and colder SSTA in the western and eastern equatorial Pacific in 2015 (Figs. 9a–c). The pattern of the SSTA difference (1997 minus 2015) in the Pacific resembles that of the central Pacific El Niño (Fig. 9c), which corresponds to stronger convection (negative OLRA) in the central-western equatorial Pacific and weaker convection (positive OLRA) in the eastern equatorial Pacific in 2015 (Figs. 9d–f). Indeed, the 2015 El Niño was identified as a “hybrid” EP–CP El Niño event by Paek et al. (2017). In 1997, enhanced convection in the eastern equatorial Pacific is associated with subsidence and thus reduced convection (positive OLRA) over the Indo-Pacific warm pool (Fig. 9d). The reduced convective heating in the eastern Indian Ocean warm pool, together with enhanced convective heating in the western tropical Indian Ocean, drives easterly wind anomalies and thus strong upwelling anomalies particularly in the EIO as discussed in section 3c(1). In contrast, during 2015 when stronger convection occurs in the central-western equatorial Pacific, the anomalous subsidence occurs in the western Pacific, the Maritime Continent, and very weakly across the equatorial Indian Ocean (Fig. 9e). This anomalous convection pattern, together with its weak amplitude, has weaker impact on the Indian Ocean equatorial wind and therefore upwelling.

Fig. 9.
Fig. 9.

SSTA (color shading; °C) and Ekman pumping velocity (contours) for November–January (NDJ) mean of (a) 1997 El Niño, (b) 2015 El Niño, and (c) the difference between them [(a) − (b)]. Units for we are 1 × 106 m s−1. The positive (negative) values are in solid black (dashed black) with values of 2, 12, 24, −24, −12, and −2. (d)–(f) As in (a)–(c), but for OLRA (W m−2) and surface wind stress.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

To further examine the impacts of CP versus EP El Niño, we extended our analysis to the 1979–2016 period when the satellite-observed OLR, ERA-Interim reanalysis wind, and ORAS4 SSH data are available. By adding 14 years (1979–92) to our record, we have added 2 EP and 2 CP El Niño events. For the entire 1979–2016 period, we have 4 EP (1982/83, 1986/87, 1997/98, and 2006/07) and 6 CP (1987/88, 1991/92, 1994/95, 2002/03, 2004/05, and 2009/10) El Niño cases (Table 1). Year 2015/16 is a hybrid CP/EP El Niño, which is discussed separately above. The composites of November–January (NDJ) mean SSHA, OLRA and surface winds for the EP and CP El Niño years show that the EP El Niños overall have stronger impacts on convection, surface wind and thus upwelling (i.e., SSHA) over the WIO and EIO especially in the EIO, compared to the CP El Niños (Fig. 10). This is because the EP El Niños are associated with enhanced convection in the central-eastern tropical Pacific, reduced convection in the Indo-Pacific warm pool, and enhanced convection in the western tropical Indian Ocean (Fig. 10d). The reduced convection in the east and enhanced convection in the west drive easterly wind anomalies and thus stronger upwelling variability over the tropical Indian Ocean. In contrast, the CP El Niños are associated with enhanced convection in the western-central tropical Pacific, weaker reduced convection in the EIO, and weaker enhanced convection in the central tropical Indian Ocean, resulting in weaker impacts on the EIO and WIO upwelling (Figs. 10b,c,e,f). This result is consistent with Kumar et al. (2006), showing that when enhanced convective heating associated with an El Niño occurs in the western-central equatorial Pacific, the anomalous subsidence (reduced convection) occurs over the Indian subcontinent, and thus may have weaker impact on atmospheric and oceanic circulations over the Indian Ocean basin.

Fig. 10.
Fig. 10.

NDJ mean SSHA (color shading; cm) and surface wind stress (arrows) for the composite of (a) all EP years, (b) all CP years, and (c) their difference [(a) − (b)] for the El Niño events since 1979 (see Table 1). (d)–(f) As in (a)–(c), but for OLRA (color shading; W m−2) and surface wind stress. SSHA data are from ORAS4 reanalysis, surface wind stress data are from the ERA-Interim reanalysis, and OLR data are from satellite observations.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

To understand whether or not the stronger impacts of the EP El Niños result from their co-occurrence with pIOD, we analyzed the composites for independent EP and CP years (Fig. 11), as well as EP–pIOD and CP–pIOD co-occurrence years (Fig. S3). Evidently, the co-occurrence of El Niño and pIOD enhances their impacts (cf. Fig. S3 and Fig. 11); however, both the independent and co-occurring years show stronger impacts of EP than CP events for the EIO and WIO. The stronger EP El Niño impacts can also be seen in Fig. 8, showing larger DLM coefficients for 1997 and 2006 EP events, and weaker coefficients for pure 2002, 2004, and 2009 CP and 2015 hybrid events. For the CP–pIOD co-occurring year of 1994, the coefficient is larger, but it is still smaller than that of the EP–pIOD co-occurring years of 1997 and 2006. Analysis of each individual CP and EP event occurring during 1979–2016 shows that even though EP El Niño events and El Niño–pIOD co-occurrence have larger impacts overall, there are special cases. For instance, although the 1994 CP–pIOD co-occurrence has apparently larger impacts than all the independent CP events, its associated Indian Ocean upwelling is comparable to that of the EP–pIOD event of 1982 (Fig. S4); and the 1986 independent EP El Niño is associated with stronger wind over the Indian Ocean than that of the 1982 EP–pIOD event. The degree of impact for each ENSO event depends on the location and strength of convection anomalies and thus wind anomalies over the Indian Ocean and Indo-Pacific warm pool region.

Fig. 11.
Fig. 11.

As in Fig. 10, but for the composites of the pure EP and pure CP El Niño years that did not co-occur with pIOD events (see Table 1 for specific years).

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

d. LOM experiments: Wave propagation and contribution of different baroclinic modes

As discussed in section 2e, since both the SLM and Bayesian DLM examine the simultaneous relationships between ENSO/IOD and upwelling, they may not fully capture the oceanic wave propagation process. In this section we use the LOM experiments to provide further insight into the wind-driven wave dynamics, confirm the results from the SLM and Bayesian DLM, and meanwhile assess the effects of wave propagation on the ENSO and IOD impacts obtained from the SLM and Bayesian DLM.

1) Model/data comparison

To verify the LOM performance in simulating the interannual variability of upwelling in the tropical Indian Ocean, we calculated the STD map of SSHA based on the LOM_MR results (Fig. 12a). Qualitatively, the LOM is able to simulate the spatial patterns of upwelling variability, with large-amplitude SSHA occurring in the SCTR region and eastern basin of the tropical Indian Ocean (Figs. 3f and 12a). Quantitatively, the LOM reasonably reproduced the observed strength of upwelling variability, albeit with weaker amplitude. For instance, the ratio of STD(lom)/STD(obs) for SSHA is 0.86, 0.66, and 0.74 and correlation for the simulated-observed SSHA is 0.90, 0.90, and 0.87 in the EIO, WIO, and Max Region, respectively. The weaker SSHA amplitudes in the LOM, especially over the WIO and Max Region that reside in the SCTR mean upwelling zone, suggest that temperature variability associated with the upwelling also plays a role in affecting SSHA, in addition to the isopycnal movement that the LOM represents. The reasonable model–data agreements suggest that the LOM is able to capture the fundamental processes governing the Indian Ocean upwelling and thus is suitable for our study. Note that the LOM cannot simulate the observed large SSH variances in the western Bay of Bengal, western Arabian Sea, and Agulhas Current region, further confirming that SSHAs in these regions result from oceanic instabilities due to nonlinearity of the oceanic system, as discussed above.

Fig. 12.
Fig. 12.

The STD of monthly SSHA (unit: cm) based on the Linear Ocean Model (LOM) (a) main run (MR), denoted by LOM_MR, (b) ENSO run (LOM_ENSO), and (c) IOD run (LOM_IOD), which are forced by the monthly CCMP2 surface wind stress from 1993 to 2016. The boxes are the same as in Figs. 3 and 5. Time series of upwelling anomalies based on AVISO SSHA (black dashed line), LOM_MR (black solid line), LOM_ENSO (red line), and LOM_IOD (blue line) averaged in the (d) EIO, (e) WIO, and (f) Max Region. The STD for each time series and the correlations between AVISO SSHA and LOM_MR, LOM_MR and LOM_ENSO, and LOM_MR and LOM_IOD are shown at the bottom of each box. For the upwelling averaged in the EIO box, the correlation between AVISO SSHA and LOM_MR is r = 0.90 (rcrit95% = 0.32), the correlation between LOM_MR and LOM_ENSO is r = 0.65 (rcrit95% = 0.34), and that between LOM_MR and LOM_IOD is r = 0.79 (rcrit95% = 0.33). For the WIO box, the correlation between AVISO SSHA and LOM_MR is r = 0.90 (rcrit95% = 0.35), the correlation between LOM_MR and LOM_ENSO is r = 0.80 (rcrit95% = 0.38), and that between LOM_MR and LOM_IOD is r = 0.65 (rcrit95% = 0.39). And for the maximum SSHA box, the correlation between AVISO SSHA and LOM_MR is r = 0.87 (rcrit95% = 0.38), the correlation between LOM_MR and LOM_ENSO is r = 0.76 (rcrit95% = 0.40), and that between LOM_MR and LOM_IOD is r = 0.65 (rcrit95% = 0.39).

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

2) Effects of climate modes

Consistent with the SLM and Bayesian DLM results, ENSO and the IOD induce similar spatial patterns of upwelling variability, and ENSO has a larger impact on upwelling than the IOD over the SCTR region (WIO and Max Region boxes; Figs. 12b,c). Both the correlation coefficients and STD values are larger for ENSO influence compared to the IOD effect (Figs. 12e,f). Consequently, ENSO has larger impacts on interannual variability of upwelling over the SCTR mean upwelling zone. In the eastern equatorial basin, ENSO and IOD have comparable effects. The SSHA correlation is 0.65 (0.79) between LOM_MR and LOM_ENSO (LOM_IOD); the STD ratio however is ~0.76 (0.68) for LOM_ENSO/LOM_MR (LOM_IOD/LOM_MR). While the IOD-induced upwelling is more in phase with the total upwelling (higher correlation), the upwelling amplitude is weaker than that induced by ENSO (smaller STD). Considering both amplitude and phase (i.e., STD and correlation), we conclude that the IOD and ENSO have comparable contributions to EIO upwelling. These results are consistent with those of the Bayesian DLM but different from those of the SLM, which shows both larger correlation and larger STD for IOD contribution than ENSO in the EIO (sections 3b and 3c). The larger correlation and STD of IOD effect in SLM led previous studies to conclude that the IOD is more important than ENSO in affecting the EIO upwelling. Since the Bayesian DLM can capture the time-varying impact of climate modes whereas the SLM cannot, the DLM provides more realistic assessment on ENSO and IOD impacts.

Remote forcing from the equator together with the longshore wind along the eastern boundary, as measured by solution (LOM_MR − LOM_DAMP) (section 2e), plays a deterministic role in causing the EIO upwelling, with the STD of SSHA being 3.29 cm, which is ~89% of the STD from the total solution (Figs. 13c,a,b). This is because equatorial winds drive eastward-propagating Kelvin waves; upon impinging on the eastern boundary, the equatorial Kelvin wave energy propagates poleward as coastal Kelvin waves and subsequently westward as Rossby waves, affecting the EIO upwelling. The effect of reflected Rossby waves however has a weaker influence on the WIO upwelling (Fig. 13d).

Fig. 13.
Fig. 13.

The STD of monthly SSHA (unit: cm) based on the LOM (a) main run (MR), denoted by LOM_MR, and (b) damp run (LOM_DAMP), which has a damping along the eastern Indian Ocean boundary from January 1993 to December 2016 and its results measure the effects of direct wind forcing in the ocean interior without the effect of eastern equatorial Indian Ocean boundary. The boxes are the same as in Figs. 3 and 5. (c),(d) Time series of upwelling anomalies based on LOM_MR (black line), LOM_DAMP (red line), and (LOM_MR − LOM_DAMP; green line) averaged in the EIO and WIO box, respectively. The correlation between LOM_MR and LOM_DAMP SSHA time series is shown at the bottom of each box, and the SSHA STDs from the MR and (LOM_MR − LOM_DAMP) are also shown.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

To confirm the Bayesian DLM results for 1997 and 2015 El Niño events, we compared the LOM solutions for these two years. We found that ENSO-related wind indeed has larger impact than that of 2015, driving stronger upwelling in the EIO and stronger downwelling in the SCTR region in 1997 (see Fig. 14). Our LOM experiments shown here are forced by CCMP2 winds (section 2e). We have compared our LOM_MR with the LOM solution forced by CCMP1 winds (Atlas et al. 2011). Their results in the open ocean are very similar, but along the east coasts, the CCMP1 winds produce stronger upwelling variability along the Sumatra and Java coasts (not shown).

Fig. 14.
Fig. 14.

OND mean maps of SSHA based on the LOM_ENSO run for (a) 1997 El Niño, (b) 2015 El Niño, and (c) the difference between them, which is calculated by subtracting 2015 values from 1997 values. The boxes are as in Fig. 3.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

3) Effect of wave propagation

To examine the effects of oceanic Rossby wave propagation on the results of SLM and Bayesian DLM, we analyzed the time evolutions of monthly SSHAs for the composites of pure El Niño years, pure pIOD years, and ENSO–IOD co-occuring years along ~5°S using AVISO data, LOM_MR, LOM_ENSO, and LOM_IOD solutions (Fig. 15). The LOM_MR well simulated the large-scale SSHA signals from the AVISO satellite observation including the westward propagation (cf. the first and second columns of Fig. 15), further demonstrating the importance of wind-driven linear wave dynamics in causing interannual SSHA over the tropical Indian Ocean. ENSO and the IOD are the dominant causes for the SSHAs during their independent years (Figs. 15b,c,f,g), and they play comparable roles for their co-occurring years across 5°S (Figs. 15j,k,o). Along 10° and 15°S ENSO effects are larger (not shown).

Fig. 15.
Fig. 15.

Longitude–time plot of monthly SSHA along ~5°S (5.25°–4.5°S average) for the composites of pure El Niño years (2002, 2004, 2009) based on (a) AVISO data, (b) LOM_MR, (c) LOM_ENSO, and (d) the SSHA due to ENSO effect (b1 × ENSO) from Bayesian DLM using the result of LOM_MR; to better reveal the propagation signals, 24-month SSHAs (January 2002–December 2003, January 2004–December 2005, and January 2009–December 2010) are shown. (e)–(h) (second row) As in (a)–(d), but for the 2012 pure positive IOD year, January 2012–December 2013. (i)–(l) (third row) As in (a)–(d), but for the composites of ENSO/IOD co-occuring years (1994, 1997, 2006 and 2015), and the results shown are AVISO SSHA, LOM_MR, LOM_ENSO, and b1 × ENSO, respectively, from January of the event year to December of the following year. (m)–(p) As in (i)–(l), but that (o) and (p) represent LOM_IOD and b1 × IOD.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

Comparing the ENSO and IOD effects from the LOM_ENSO and LOM_IOD experiments, which fully capture the effects of oceanic wave propagation, with the b1 × ENSO and b1 × IOD terms of the Bayesian DLM for the SSHA from LOM_MR (third and fourth columns of Fig. 15), it is evident that the westward propagations are not properly captured in the Bayesian DLM. For instance, the westward propagations of positive SSHAs from September to March of pure ENSO years, June to December of pure pIOD, and July to March of the composites for co-occurring years (Fig. 15, second and third columns) are not shown in the Bayesian DLM (Fig. 15, fourth column).

Quantitatively, the propagations of oceanic waves have significant contributions to the upwelling especially in the western basin (Fig. 16). In the EIO, the area-averaged SSHAs from the LOM_ENSO experiment and from b1 × ENSO of the Bayesian DLM are similar, with a correlation coefficient of 0.96 and STD ratio = STD(b1 × ENSO)/STD(LOM_ENSO) of 1 from 1993 to 2016 (Figs. 16a,b). Similarly, the correlation for LOM_IOD and b1 × IOD is 0.95 and has an STD ratio of 0.96. This is because the eastward propagation speeds of equatorial Kelvin waves are fast. It takes only ~1 month (1.5 months) for the Kelvin waves associated with the first (second) baroclinic mode to propagate across the equatorial basin. Consequently, the monthly SSHA and climate mode indices can well capture the equatorial Kelvin waves’ effects associated with the first two baroclinic modes, which dominate the SSHAs in the tropical Indian Ocean (see below).

Fig. 16.
Fig. 16.

Time series of the monthly anomalous upwelling, that is, SSHA (cm) based on LOM_ENSO experiment (black line) and the ENSO effect from Bayesian DLM (red line) averaged in the (a) EIO, (c) WIO, (e) Max Region, and (g) boxed region of (2°–15°S, 50°–65°E) within the WIO box for the period of 1993 to 2016. (b),(d),(f),(h) As in (a),(c),(e),(g), respectively, but for the SSHA from LOM_IOD experiment (black line) and the IOD effect from Bayesian DLM (red line).

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

In contrast, the Bayesian DLM cannot accurately capture the ENSO and IOD effects in the western basin, and the errors increase progressively toward the west due to the westward propagation of oceanic Rossby waves, whose speeds are much slower than that of the equatorial Kelvin waves. The correlations between the ENSO-induced SSHAs from b1 × ENSO and LOM_ENSO are 0.8 in the Max Region in the central basin and 0.72 in the WIO, and the corresponding ratios of STD (b1 × ENSO)/STD(LOM_ENSO) are 68% and 57% in the two regions (Figs. 16e,c). The correlations and STD ratios for the IOD show similar deteriorations toward the west, with r = 0.78 and 0.69 and STD ratios = 58% and 50% in the Max and WIO regions, respectively (Figs. 16f,d). In a smaller region of 15°–2°S, 50°–65°E, the correlation coefficient for b1 × ENSO and LOM_ENSO SSHAs is 0.64 and their STD ratio is 48%, and those for the IOD are 0.6 and 39% (Figs. 16g,h). These results suggest that even though the conclusions regarding the relative importance of ENSO and IOD from the LOM agree with those of Bayesian DLM, neglecting wave propagation in the DLM causes quantitative errors in assessing ENSO and IOD impacts on sea level in the western Indian Ocean basin.

4) Contributions from lower-order baroclinic modes

Existing studies demonstrated that the first two baroclinic modes can well explain the seasonal to interannual variability of zonal surface current in the equatorial Indian Ocean (e.g., Han et al. 1999; Nagura and McPhaden 2010). For the variability of Equatorial Undercurrent, however, intermediate modes (3–8) also have important contributions (Chen et al. 2015). To explore how many baroclinic modes are required to properly simulate the interannual variability of Indian Ocean upwelling (i.e., SSHA), we plotted out the STD maps of SSHA contributed from different baroclinic modes. The first two modes explain a large fraction of the basinwide SSHA, with their sum explaining 83%, 92%, and 93% SSHA STD of the LOM_MR in the EIO, Max Region, and WIO, respectively and the second baroclinic mode having a larger effect than the first mode. The first three modes together account for 91%, 96%, and 97% in the three regions and the third mode has its largest contribution in the EIO (Figs. 17a–d,f). The sum of the first eight baroclinic modes explains ~99%, 100%, and 100% in EIO, Max Region, and WIO. The low-order modes (1–2) have faster propagation speeds and thus experience weaker damping by friction, explaining 83%–93% SSHA in the three regions. The intermediate modes (3–8) subject to larger friction due to their relatively slower speeds; however, they also contribute 16%, 8%, and 7% of upwelling (SSHA) in the EIO, Max Region, and WIO. The high-order modes (modes 9–25), by contrast, are subject to large friction and they are essentially in local “pseudo-Ekman balance” (Han 2005); therefore, their associated pressure gradients and SSHA are negligible (Fig. 17g).

Fig. 17.
Fig. 17.

The STD maps of monthly SSHA (cm) based on the sum of (a) the first 25 baroclinic modes, (b) the first two baroclinic modes, (c) the first three baroclinic modes, (e) mode 4 to mode 8, and (g) mode 9 to mode 25 of the LOM_MR result. (d),(f),(h) The contributions for modes 1, 2, and 3, respectively. The boxes are the same as those of Fig. 3.

Citation: Journal of Climate 33, 4; 10.1175/JCLI-D-19-0386.1

4. Summary and discussion

In this paper, we investigated the interannual variability of upwelling in the tropical Indian Ocean over the SCTR mean upwelling zone and the seasonal upwelling area in the EIO where interannual variations are strong (Figs. 1 and 3). The approach is to carry out observational analyses using monthly satellite, in situ, and reanalysis datasets and perform ocean model experiments using a linear continuously stratified ocean model (LOM), whose solutions were found for each baroclinic mode and the total solution is represented by the sum of the first 25 modes (section 2e). For the observational analyses, we applied both the conventional SLM and the relatively new Bayesian DLM. Compared to the SLM, the DLM has the advantage of capturing the time-varying impacts of ENSO and IOD, the two dominant interannual climate modes over the tropical Indo-Pacific basin (sections 2bd). The major period of interest is 1993–2016, when satellite SSH data are available. To assess the impacts of EP versus CP El Niño events, our analysis also goes back to 1979, when satellite observed OLR, ERA-Interim wind, and ORAS4 SSH reanalysis data are available (section 2a).

Consistent with existing studies, larger interannual variability of upwelling is observed over the EIO and SCTR region, with the maximum shoaling of D20 being ~50 m and SSHA dropping of ~20 cm averaged over the EIO (Fig. 4), the eastern node of the IOD, in the fall of 1997 when El Niño and pIOD co-occurred (Fig. 2). Two areas over the SCTR were chosen to carry out time series analysis, the WIO that covers the area of shallowest mean thermocline and the Max Region of the eastern SCTR, where the STD of D20A obtains its maximum of ~10 m for the 1993–2016 period (Figs. 3 and 4). Interannual variability of upwelling averaged in the WIO (or Max Region) and EIO are negatively correlated (Figs. 4 and 7), attaining maximum correlation of r = −0.73 when the WIO lags the EIO by 2 months using AVISO SSHA because of the westward propagation of Rossby waves generated in the eastern basin (Figs. 5a and 15). Various upwelling indicators have been tested (e.g., D20A, SSHA, SSTA). In the tropical Indian Ocean, baroclinic response dominates barotropic response to surface wind forcing; therefore, SSHA mirrors D20A and can serve as a good indicator for upwelling (Fig. 4). Since the AVISO satellite SSHA has higher spatial resolution, we chose to use AVISO SSHA to represent upwelling for examining the ENSO and IOD impacts.

For the upwelling over the SCTR region, both ENSO and the IOD are important in causing the interannual variability, but ENSO plays a more important role than the IOD in the WIO and Max Region (Figs. 3 and 5), as measured by the higher correlation and larger STD ratio against satellite observed upwelling (i.e., SSHA). This conclusion holds for both the SLM and Bayesian DLM (Fig. 6). The Bayesian DLM consistently outperforms the SLM in assessing the ENSO and IOD impacts, with correlation coefficient between observed and ENSO-contributed SSHA being r = 0.79 (0.62) in the WIO and r = 0.83 (0.65) in the Max Region using the Bayesian DLM (SLM). For the EIO upwelling, SLM results show a somewhat larger effect of the IOD (r = 0.68 and STD ratio = 0.67) than ENSO (r = 0.56 and STD ratio = 0.56), as concluded in previous studies that also used the SLM, whereas Bayesian DLM shows comparable effects. While ENSO-contributed upwelling in Bayesian DLM has a slightly lower correlation with the observed upwelling (r = 0.70) compared to the IOD (r = 0.79), the STD ratio is somewhat larger for ENSO (0.79) relative to that of the IOD (0.72). Given the better performance of the Bayesian DLM, we conclude that ENSO and the IOD have comparable influences on the EIO upwelling. This conclusion is further confirmed by the LOM experiments (Figs. 12 and 15). Therefore, the conclusion of stronger ENSO (IOD) impacts south (north) of 10°S from previous studies (section 1) may hold for their analysis periods, but does not hold for the 1993–2016 period examined here based on the Bayesian DLM and LOM experiments. While both ENSO and the IOD contribute to the strong upwelling variability in the EIO during summer and fall, ENSO is the predominant cause for the strong variability of upwelling over the SCTR region during winter and spring. The two climate modes together explain the majority of observed seasonality of upwelling in the tropical Indian Ocean (Fig. 5).

ENSO and the IOD cause similar upwelling patterns, with El Niño and pIOD inducing anomalous upwelling (downwelling) in the eastern tropical basin (SCTR region) (Fig. 7 and Figs. S1 and S2 in the online supplemental material). The anomalous southeasterly and equatorial easterly winds associated with the El Niño and pIOD force coastal and equatorial upwelling, with largest amplitude occurring in the eastern basin due to eastward propagation of equatorial Kelvin waves and subsequent poleward propagation of coastally trapped waves (Fig. 13). Meanwhile, the off-equatorial negative Ekman pumping velocities deepen the thermocline and raise sea level. These signals propagate westward as Rossby waves, reducing the mean upwelling in the SCTR region (Fig. 7). Therefore, El Niño and pIOD cause out-of-phase relationships between EIO and WIO upwelling variability, and upwelling anomalies generally intensify when El Niño and pIOD co-occur (Figs. 2, 4, and 6). The strong variability of upwelling in the Max Region results largely from ENSO, and to a lesser degree from the IOD, because the amplitudes of D20A and SSHA are roughly proportional to the zonally integrated effect of we, which obtains the maximum value in this area [Fig. 7; section 3c(1)].

In general, eastern Pacific El Niño events have stronger impacts on Indian Ocean upwelling particularly in the EIO, compared to the central Pacific El Niños and the 2015 hybrid El Niño for the 1993–2016 and 1979–2016 periods (Figs. 811 and Fig. S3). This is because the EP El Niño events are associated with more reduced convection (positive OLRA) in the EIO and more enhanced convection (negative OLRA) in the WIO basin, compared to the CP and hybrid El Niños. The reduced convection in the east and enhanced convection in the west drive easterly wind anomalies and thus stronger upwelling variability. While the EP El Niño and pIOD co-occurrences overall intensify the EIO and WIO upwelling anomalies, the pure EP and CP El Niño events (independent of pIOD) also show stronger impacts of EP El Niños (Fig. 11; Fig. S3). Analysis of each individual CP and EP event during 1979–2016, however, suggests that there are special cases. The surface wind and SSHA associated with the 1994 CP–pIOD event are similar to those of the 1982 EP–pIOD event (Fig. S4), and the 1986 independent EP El Niño is associated with similar or even stronger surface wind and convection than the 1982 EP–pIOD event. The sensitivity of anomalous Indian Ocean wind and convection to the locations of convection anomalies of the Pacific associated with El Niños raises the challenge for predicting ENSO impacts on Indian Ocean upwelling. Note that our conclusions are based on the 1993–2016 and 1979–2016 periods. It is possible that the impacts of EP versus CP El Niños on interannual variability of IO upwelling experience interdecadal variability, due to the influences of decadal climate modes (e.g., interdecadal Pacific oscillation). This is an interesting area for our future study.

Even though the Bayesian DLM shows significant advantages over the SLM in assessing the climate modes’ impacts, the simultaneous relationships between the monthly climate mode indices and upwelling indicators, as shown by Eqs. (2) and (3), cannot properly capture the effects of oceanic Rossby wave propagation, which is much slower than the monthly time scale of the data. Indeed, the hierarchy of carefully designed LOM experiments confirms the major conclusions drawn by the Bayesian DLM (Fig. 12). Quantitatively however the LOM solutions demonstrated large effects of Rossby waves in the western basin, which were not captured by the Bayesian DLM (Figs. 15 and 16). The correlation between the LOM-simulated and DLM-estimated SSHA is only 0.64 and the STD ratio (DLM/LOM) is 48% for ENSO, and 0.6 and 39% for IOD in the region 15°–2°S, 50°–65°E in the western Indian Ocean. While the first two baroclinic modes explain 83%, 92%, and 93% of the upwelling variability in the EIO, Max Region, and WIO, respectively, the intermediate modes (3–8) are needed to fully represent upwelling in the tropical Indian Ocean (Fig. 17).

Because the Indian Ocean upwelling directly affects biogeochemical processes (e.g., nutrients, CO2 flux at the air–sea interface, etc.) and is tied to thermocline variability, upper ocean heat content, and SSTA, results from this study have important implications for understanding variability of marine ecosystems and climate. Careful studies are needed to systematically investigate the cross-disciplinary impacts of Indian Ocean upwelling.

Acknowledgments

We gratefully acknowledge the support from the National Aeronautics and Space Administration (NASA) Ocean Surface Topography Science Team Award NNX17AI63G and NASA Physical Oceanography Program Award NNX17AH25G. CCMP version-2.0 vector wind analyses are produced by Remote Sensing Systems. Data are available at www.remss.com.

APPENDIX

The Linear Continuously Stratified Ocean Model

As described in McCreary et al. 1996 (Han 2005; Han et al. 2011), the equations of motion are linearized about a background state of rest with stratification represented by Brunt–Väisälä frequency, and the ocean bottom is assumed flat at D = 4000 m. Under the above approximations, the solutions for the zonal velocity u, meridional velocity υ, and pressure p can be expanded in the vertical normal modes with eigenfunctions Fn(z):
u=n=0NunFn,
υ=n=0NυnFn,and
p=n=0NpnFn,
where N is the total mode number. The terms un, υn, and pn are expansion coefficients that satisfy the following governing equations:
(t+Acn2)unfυn+1ρ¯pnx=τxZnρ¯Hn+ϑn2un,
(t+Acn2)υn+fun+1ρ¯pny=τyZnρ¯Hn+ϑn2υn,and
(t+Acn2)pnρ¯cn2+unx+υny=0.
In the above, cn represents the equatorial Kelvin wave speed for vertical mode number n. The cn values for the first eight baroclinic modes (n = 1, 2, …, 8) are 2.64, 1.67, 1.05, 0.75, 0.60, 0.49, 0.42, and 0.37 m s−1 with observed stratification within the Indian Ocean (e.g., McCreary et al. 1996; Han et al. 2004). The factors Zn=D0Z(z)Fndz and Hn=D0Fn2dz determine how strongly the driving wind couples to each mode; f = βy is the Coriolis parameter under equatorial β-plane approximation; and Z(z) is the vertical profile of wind that is introduced as a body force, where Z(z) is constant in the upper 50 m and linearly decreases to zero from 50- to 100-m depth. The terms associated with A/cn2 represent vertical friction, with A = 0.000 13 cm2 s−3. Density ρ¯=1gcm3 is a typical density value of seawater. More details can be found in McCreary et al. (1996).

REFERENCES

  • Annamalai, H., P. Liu, and S.-P. Xie, 2005: Southwest Indian Ocean SST variability: Its local effect and remote influence on Asian monsoons. J. Climate, 18, 41504167, https://doi.org/10.1175/JCLI3533.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ashok, K., W. L. Chan, T. Motoi, and T. Yamagata, 2004: Decadal variability of the Indian Ocean dipole. Geophys. Res. Lett., 31, L24207, https://doi.org/10.1029/2004GL021345.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Atlas, R., R. N. Hoffman, J. Ardizzone, S. M. Leidner, J. C. Jusem, D. K. Smith, and D. Gombos, 2011: A cross-calibrated, multiplatform ocean surface wind velocity product for meteorological and oceanographic applications. Bull. Amer. Meteor. Soc., 92, 157174, https://doi.org/10.1175/2010BAMS2946.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Balmaseda, M. A., K. E. Trenberth, and E. Källén, 2013: Distinctive climate signals in reanalysis of global ocean heat content. Geophys. Res. Lett., 40, 17541759, https://doi.org/10.1002/grl.50382.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Baquero-Bernal, A., and M. Latif, 2005: Wind-driven oceanic Rossby waves in the tropical South Indian Ocean with and without an active ENSO. J. Phys. Oceanogr., 35, 729746, https://doi.org/10.1175/JPO2723.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Birol, F., and R. Morrow, 2001: Source of the baroclinic waves in the southeast Indian Ocean. J. Geophys. Res., 106, 91459160, https://doi.org/10.1029/2000JC900044.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cai, W., and et al. , 2018: Increased variability of eastern Pacific El Niño under greenhouse warming. Nature, 564, 201206, https://doi.org/10.1038/s41586-018-0776-9.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chambers, D. P., B. D. Tapley, and R. H. Stewart, 1999: Anomalous warming in the Indian Ocean coincident with El Niño. J. Geophys. Res., 104, 30353047, https://doi.org/10.1029/1998JC900085.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chelliah, M., and P. Arkin, 1992: Large-scale interannual variability of monthly outgoing longwave radiation anomalies in the global tropics. J. Climate, 5, 371389, https://doi.org/10.1175/1520-0442(1992)005<0371:LSIVOM>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, G., W. Han, Y. Li, D. Wang, and M. McPhaden, 2015: Seasonal-to-interannual time scale dynamics of the Equatorial Undercurrent in the Indian Ocean. J. Phys. Oceanogr., 45, 15321553, https://doi.org/10.1175/JPO-D-14-0225.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, G., W. Han, Y. Li, and D. Wang, 2016: Interannual variability of equatorial eastern Indian Ocean upwelling: Local versus remote forcing. J. Phys. Oceanogr., 46, 789807, https://doi.org/10.1175/JPO-D-15-0117.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dee, D., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, https://doi.org/10.1002/qj.828.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Deepa, J. S., C. Gnanaseelan, R. Kakatkar, A. Parekh, and J. S. Chowdary, 2018: The interannual sea level variability in the Indian Ocean as simulated by an ocean general circulation model. Int. J. Climatol., 38, 11321144, https://doi.org/10.1002/joc.5228.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ebisuzaki, W., 1997: A method to estimate the statistical significance of a correlation when the data are serially correlated. J. Climate, 10, 21472153, https://doi.org/10.1175/1520-0442(1997)010<2147:AMTETS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fu, L. L., and R. D. Smith, 1996: Global ocean circulation from satellite altimetry and high resolution computer simulation. Bull. Amer. Meteor. Soc., 77, 26252636, https://doi.org/10.1175/1520-0477(1996)077<2625:GOCFSA>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fukumori, I., R. Raghunath, and L.-L. Fu, 1998: Nature of global large-scale sea level variability in relation to the atmospheric forcing: A modeling study. J. Geophys. Res., 103, 54935512, https://doi.org/10.1029/97JC02907.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gill, E. A., 1982: Atmosphere–Ocean Dynamics. Academic Press, 662 pp.

  • Gnanaseelan, C., and B. H. Vaid, 2010: Interannual variability in the biannual Rossby waves in the tropical Indian Ocean and its relation to Indian Ocean dipole and El Niño forcing. Climate Dyn., 60, 2740, https://doi.org/10.1007/s10236-009-0236-z.

    • Search Google Scholar
    • Export Citation
  • Han, W., 2005: Origins and dynamics of the 90-day and 30-60-day variations in the equatorial Indian Ocean. J. Phys. Oceanogr., 35, 708728, https://doi.org/10.1175/JPO2725.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., and P. J. Webster, 2002: Forcing mechanisms of sea-level interannual variability in the Bay of Bengal. J. Phys. Oceanogr., 32, 216239, https://doi.org/10.1175/1520-0485(2002)032<0216:FMOSLI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., J. P. McCreary, D. L. T. Anderson, and A. J. Mariano, 1999: Dynamics of the eastward surface jets in the equatorial Indian Ocean. J. Phys. Oceanogr., 29, 21912209, https://doi.org/10.1175/1520-0485(1999)029<2191:DOTESJ>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., P. J. Webster, R. Lukas, P. Hacker, and A. Hu, 2004: Impact of atmospheric intraseasonal variability in the Indian Ocean: Low-frequency rectification in equatorial surface current and transport. J. Phys. Oceanogr., 34, 13501372, https://doi.org/10.1175/1520-0485(2004)034<1350:IOAIVI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., J. P. McCreary, Y. Masumoto, J. Vialard, and B. Duncan, 2011: Basin resonances in the equatorial Indian Ocean. J. Phys. Oceanogr., 41, 12521270, https://doi.org/10.1175/2011JPO4591.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., G. Meehl, D. Stammer, A. Hu, B. Hamlington, J. Kenigson, H. Palanisamy, and P. Thompson, 2017a: Spatial patterns of sea level variability associated with natural internal climate modes. Surv. Geophys., 38, 217250, https://doi.org/10.1007/s10712-016-9386-y.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., G. Meehl, A. Hu, J. Zheng, J. Kenigson, J. Vialard, B. Rajagopalan, and Yanto, 2017b: Decadal variability of Indian and Pacific Walker cells: Do they co-vary on decadal timescales? J. Climate, 30, 84478468, https://doi.org/10.1175/JCLI-D-16-0783.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Han, W., D. Stammer, G. A. Meehl, A. Hu, F. Sienz, and L. Zhang, 2018: Multi-decadal trend and decadal variability of the regional sea level over the Indian Ocean since the 1960s: Roles of climate modes and external forcing. Climate, 6, 51, https://doi.org/10.3390/cli6020051.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hermes, J. C., and C. J. C. Reason, 2008: Annual cycle of the South Indian Ocean (Seychelles-Chagos) thermocline ridge in a regional ocean model. J. Geophys. Res., 113, C04035, https://doi.org/10.1029/2007JC004363.

    • Search Google Scholar
    • Export Citation
  • Hosoda, S., T. Ohira, and T. Nakamura, 2008: A monthly mean dataset of global oceanic temperature and salinity derived from Argo float observations. JAMSTEC Rep. Res. Dev., 8, 4759, https://doi.org/10.5918/jamstecr.8.47.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hu, S., and et al. , 2019: Interannual to decadal variability of upper-ocean salinity in the southern Indian Ocean and the Role of the Indonesian Throughflow. J. Climate, 32, 64036421, https://doi.org/10.1175/JCLI-D-19-0056.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huang, B., and J. L. Kinter III, 2002: Interannual variability in the tropical Indian Ocean. J. Geophys. Res., 107, 3199, https://doi.org/10.1029/2001JC001278.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Izumo, T., and et al. , 2010: Influence of the state of the Indian Ocean Dipole on the following year’s El Niño. Nat. Geosci., 3, 168172, https://doi.org/10.1038/ngeo760.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Jury, M. R., and B. Huang, 2004: The Rossby wave as a key mechanism of Indian Ocean climate variability. Deep-Sea Res. I, 51, 21232136, https://doi.org/10.1016/j.dsr.2004.06.005.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Klein, S. A., B. J. Soden, and N.-C. Lau, 1999: Remote sea surface variations during ENSO: Evidence for a tropical atmospheric bridge. J. Climate, 12, 917932, https://doi.org/10.1175/1520-0442(1999)012<0917:RSSTVD>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Krishnaswamy, J., S. Vaidyanathan, B. Rajagopalan, M. Bonell, M. Sankaran, R. S. Bhalla, and S. Badiger, 2015: Non-stationary and non-linear influence of ENSO and Indian Ocean dipole on the variability of Indian monsoon rainfall and extreme rain events. Climate Dyn., 45, 175184, https://doi.org/10.1007/s00382-014-2288-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kumar, K. K., B. Rajagopalan, and M. A. Cane, 1999: On the weakening relationship between the Indian monsoon and ENSO. Science, 284, 21562159, https://doi.org/10.1126/science.284.5423.2156.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kumar, K. K., B. Rajagopalan, M. Hoerling, G. Bates, and M. Cane, 2006: Unraveling the mystery of Indian monsoon failure during El Niño events. Science, 314, 115119, https://doi.org/10.1126/science.1131152.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Levitus, S., and et al. , 2012: World ocean heat content and thermosteric sea level change (0–2000 m), 1955–2010. Geophys. Res. Lett., 39, L10603, https://doi.org/10.1029/2012GL051106.

    • Search Google Scholar
    • Export Citation
  • Li, Y., and W. Han, 2015: Decadal sea level variations in the Indian Ocean investigated with HYCOM: Roles of climate modes, ocean internal variability and stochastic wind forcing. J. Climate, 28, 91439165, https://doi.org/10.1175/JCLI-D-15-0252.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Masumoto, Y., and G. Meyers, 1998: Forced Rossby waves in the southern tropical Indian Ocean. J. Geophys. Res., 103, 27 58927 602, https://doi.org/10.1029/98JC02546.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., 1981: A linear stratified ocean model of the coastal undercurrent. Philos. Trans. Roy. Soc. London, 302A, 385413, https://doi.org/10.1098/rsta.1981.0176.

    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., P. K. Kundu, and R. L. Molinari, 1993: A numerical investigation of dynamics, thermodynamics and mixed-layer processes in the Indian Ocean. Prog. Oceanogr., 31, 181244, https://doi.org/10.1016/0079-6611(93)90002-U.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCreary, J. P., W. Han, D. Shankar, and S. R. Shetye, 1996: Dynamics of the East India Coastal Current: 2. Numerical solutions. J. Geophys. Res., 101, 13 99314 010, https://doi.org/10.1029/96JC00560.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Murtugudde, R., S. Signorini, J. Christian, A. Busalacchi, C. McClain, and J. Picaut, 1999: Ocean color variability of the tropical Indo-Pacific basin observed by SeaWiFS during 1997-98. J. Geophys. Res., 104, 18 35118 366, https://doi.org/10.1029/1999JC900135.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Murtugudde, R., J. P. McCreary, and A. J. Busalacchi, 2000: Oceanic processes associated with anomalous events in the Indian Ocean with relevance to 1997–1998. J. Geophys. Res., 105, 32953306, https://doi.org/10.1029/1999JC900294.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nagura, M., and M. J. McPhaden, 2010: Dynamics of zonal current variations associated with the Indian Ocean dipole. J. Geophys. Res., 115, C11026, https://doi.org/10.1029/2010JC006423.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Paek, H., J.-Y. Yu, and C. Qian, 2017: Why were the 2015/2016 and 1997/1998 extreme El Niños different? Geophys. Res. Lett., 44, 18481856, https://doi.org/10.1002/2016GL071515.

    • Search Google Scholar
    • Export Citation
  • Périgaud, C., and P. Delecluse, 1992: Annual sea level variations in the southern tropical Indian Ocean from Geosat and shallow-water simulations. J. Geophys. Res., 97, 20 16920 178, https://doi.org/10.1029/92JC01961.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Périgaud, C., and P. Delecluse, 1993: Interannual sea level variations in the tropical Indian Ocean from Geosat and shallow water simulations. J. Phys. Oceanogr., 23, 19161934, https://doi.org/10.1175/1520-0485(1993)023<1916:ISLVIT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Petris, G., 2010: dlm: Bayesian and likelihood analysis of dynamic linear models. R package version 1.1-1, https://rdrr.io/cran/dlm/.

  • Petris, G., S. Petrone, and P. Campagnoli, 2009: Dynamic Linear Models with R. Springer, 252 pp (see especially pp. 31–84).

    • Crossref
    • Export Citation
  • Potemra, J. T., 2001: Contribution of equatorial Pacific winds to southern tropical Indian Ocean Rossby waves. J. Geophys. Res., 106, 24072422, https://doi.org/10.1029/1999JC000031.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rao, S. A., and S. K. Behera, 2005: Subsurface influence on SST in the tropical Indian Ocean: Structure and interannual variability. Dyn. Atmos. Oceans, 39, 103135, https://doi.org/10.1016/j.dynatmoce.2004.10.014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, https://doi.org/10.1029/2002JD002670.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • R Development Core Team, 2016: R: A language and environment for statistical computing. R Foundation for Statistical Computing, https://www.R-project.org/.

  • Saji, N. H., B. N. Goswami, P. N. Vinayachandran, and T. Yamagata, 1999: A dipole mode in the tropical Indian Ocean. Nature, 401, 360363, https://doi.org/10.1038/43854.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schott, F. A., S.-P. Xie, and J. P. McCreary, 2009: Indian Ocean circulation and climate variability. Rev. Geophys., 47, RG1002, https://doi.org/10.1029/2007RG000245.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shinoda, T., H. H. Hendon, and M. A. Alexander, 2004: Surface and subsurface dipole variability in the Indian Ocean and its relation with ENSO. Deep-Sea Res. I, 51, 619635, https://doi.org/10.1016/j.dsr.2004.01.005.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Susanto, R. D., A. L. Gordon, and Q. N. Zheng, 2001: Upwelling along the coasts of Java and Sumatra and its relation to ENSO. Geophys. Res. Lett., 28, 15991602, https://doi.org/10.1029/2000GL011844.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Takahashi, K., A. Montecinos, K. Goubanova, and B. Dewitte, 2011: ENSO regimes: Reinterpreting the canonical and Modoki El Niño. Geophys. Res. Lett., 38, L10704, https://doi.org/10.1029/2011GL047364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Timmermann, A., and et al. , 2018: El Niño–Southern Oscillation complexity. Nature, 559, 535545, https://doi.org/10.1038/s41586-018-0252-6.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tozuka, T., T. Yokoi, and T. Yamagata, 2010: A modeling study of interannual variations of the Seychelles Dome. J. Geophys. Res., 115, C04005, https://doi.org/10.1029/2009JC005547.

    • Search Google Scholar
    • Export Citation
  • Trenary, L., and W. Han, 2012: Intraseasonal to interannual variability of South Indian Ocean sea level and thermocline: Remote versus local forcing. J. Phys. Oceanogr., 42, 602627, https://doi.org/10.1175/JPO-D-11-084.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wang, L., C. J. Koblinsky, and S. Howden, 2001: Annual Rossby wave in the Southern Indian Ocean: Why does it “appear” to break down in the middle ocean? J. Phys. Oceanogr., 31, 5474, https://doi.org/10.1175/1520-0485(2001)031<0054:ARWITS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 356360, https://doi.org/10.1038/43848.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., J. Scott, R. Hoffman, M. Leidner, R. Atlas, and J. Ardizzone, 2015: Remote Sensing Systems Cross-Calibrated Multi-Platform (CCMP) 6-hourly ocean vector wind analysis product on 0.25 deg grid, version 2.0. Remote Sensing Systems, accessed 2019, www.remss.com/measurements/ccmp.

  • White, W. B., 2001: Evidence for coupled Rossby waves in the annual cycle of the Indo-Pacific Ocean. J. Phys. Oceanogr., 31, 29442957, https://doi.org/10.1175/1520-0485(2001)031<2944:EFCRWI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Woodberry, K. E., M. E. Luther, and J. J. O’Brien, 1989: The wind-driven seasonal circulation in the southern tropical Indian Ocean. J. Geophys. Res., 94, 17 98518 002, https://doi.org/10.1029/JC094Ic12P17985.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xie, S.-P., and P. A. Arkin, 1996: Analysis of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9, 840858, https://doi.org/10.1175/1520-0442(1996)009<0840:AOGMPU>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xie, S.-P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78, 25392558, https://doi.org/10.1175/1520-0477(1997)078<2539:GPAYMA>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xie, S.-P., H. Annamalai, F. A. Schott, and J. P. McCreary, 2002: Structure and mechanisms of south Indian Ocean climate variability. J. Climate, 15, 864878, https://doi.org/10.1175/1520-0442(2002)015<0864:SAMOSI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 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, 33523368, https://doi.org/10.1175/2010JCLI3429.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yang, J., L. Yu, C. J. Koblinsky, and D. Adamec, 1998: Dynamics of the seasonal variations in the Indian Ocean from TOPEX/POSEIDON sea surface height and an ocean model. Geophys. Res. Lett., 25, 19151918, https://doi.org/10.1029/98GL01401.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yokoi, T., T. Tozuka, and T. Yamagata, 2008: Seasonal variation of the Seychelles Dome. J. Climate, 21, 37403754, https://doi.org/10.1175/2008JCLI1957.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yokoi, T., T. Tozuka, and T. Yamagata, 2009: Seasonal variation of the Seychelles Dome simulated in the CMIP3 models. J. Phys. Oceanogr., 39, 449457, https://doi.org/10.1175/2008JPO3914.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yu, J.-Y., Y. Zou, S. T. Kim, and T. Lee, 2012: The changing impact of El Niño on US winter temperatures. Geophys. Res. Lett., 39, L15702, https://doi.org/10.1029/2012GL052483.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yu, W., B. Xiang, L. Liu, and N. Liu, 2005: Understanding the origins of interannual thermocline variations in the tropical Indian Ocean. Geophys. Res. Lett., 32, L24706, https://doi.org/10.1029/2005GL024327.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhuang, W., M. Feng, Y. Du, A. Schiller, and D. Wang, 2013: Low-frequency sea level variability in the southern Indian Ocean and its impacts on the oceanic meridional transports. J. Geophys. Res. Oceans, 118, 13021315, https://doi.org/10.1002/jgrc.20129.

    • Crossref
    • Search Google Scholar
    • Export Citation

Supplementary Materials

Save