Fig. 1
The boxes enclose some regions that are important for the definition of the indexes used in this work. The Niño3.4 Index was produced by averaging the SST in the Niño3.4 region. The South Tropical Zonal Gradient (STZG) and the North Tropical Zonal Gradient (NTZG) Indexes are defined as averages of SST anomalies in the region 6 (region 7) minus those in the region 5 (region 8), respectively.
Table 1. Acronyms and specifications of some tropical Climate Indexes and some of the Climate Fields used
SOISouthern Oscillation IndexStandardised surface level pressure differences between Tahiti and Darwin (Trenberth, 1984).Niño3.4Niño3.4Average anomalies of sea surface temperature in the Niño3.4 region (170°W–120°W, 5°S–5°N, Trenberth, 1997).WWV-westWarm Water Volume-westThe integrated warm water volume (WWV) above the 20°C isotherm between 5°N and 5°S, 120°E to 155°W (Meinen and McPhaden, 2000) obtained from Australian Bureau of Meteorology Research Center (BMRC) reanalysis.WWV-eastWarm Water Volume-eastThe integrated warm water volume (WWV) above the 20°C isotherm between 5°N–5°S, 155°W and 80°W (Meinen and McPhaden, 2000) obtained from Australian Bureau of Meteorology Research Center (BMRC) reanalysis.TSATropical Southern AtlanticAnomaly of the average of the monthly SST between (0°S–20°S, 10°E–30°W). HadI SST and NOAA OI 1°×1° data sets are used to create this index (Enfield et al., 1999).PMMPacific Meridional Mode SSTThe Pacific Meridional Mode is defined via applying maximum co-variance analysis (MCA) to sea surface temperature (SST) and the zonal and meridional components of the 10 m wind field over the time period 1950–2005, from the NCEP/NCAR reanalysis. To define the spatial pattern, data are defined over the region (21°S–32°N, 74°W–15°E), and spatially smoothed (three longitude by two latitude points, (Chiang and Vimont, 2004)).IODIndian Ocean Dipole ModeThe Indian Ocean Dipole is defined as the SST anomaly difference between the western equatorial Indian Ocean (10°S–10°N, 50°E–70°E) and the south eastern equatorial Indian Ocean (10°S–0°, 90°E–110°E), sometimes referred to as Dipole Mode Index (DMI), Saji et al. (1999).SSTSea Surface TemperatureSea Surface Temperature from the ERA-Interim reanalysis (Dee et al., 2011). 0.75°×0.75° resolution (www.ecmwf.int/en/forecast/datasets).MTTMiddle Tropospheric TemperatureMiddle Tropospheric Temperature data from NOAA's TIROS-N polar orbiting satellites and adjusted for time-dependent biases by the Global Hydrology and Climate Center at the University of Alabama in Huntsville (UAH). 2.5°×2.5° resolution (www.ncdc.noaa.gov/temp-and-precip/msu/index.php).NTZGNorth Tropical Zonal GradientSST anomaly differences between (5°–15°N, 160°E–170°W) and (5°–15°N, 120°W–90°W) regions.STZGSouth Tropical Zonal GradientSST anomaly differences between (5°–15°S, 160°E–170°W) and (5°–15°S, 120°W–90°W) regions.ISVIntraSeasonal VariabilityIndex of high-frequency surface wind forcing in the western equatorial Pacific bandpass filtered at periods of 30–95 d (5°S–5°N, 120°E–180°E, McPhaden et al., 2006).
Fig. 2
Seasonal standard deviation values of the Niño3.4 Index (black squares), the SOI (orange circles), the WWV Index (red triangles) and the TSA Index (blue diamonds).
Fig. 3
The lead–lag correlation coefficients between the Niño3.4 Index and the WWV Index (long dashed line), and between the former index and the third PC (short dashed line) of the MTT anomalies in the GSE region are depicted in (a) (summer) and (b) (autumn). In the background, we have represented the autocorrelation functions of the Niño3.4 Index (black solid line). The red dashed lines represent the statistical significance threshold at 95% confidence level.
Fig. 4
The RMSE of the NW model hindcast of the Niño3.4 Index (dashed line) and the RMSE of the NSW model hindcast (dot–dashed line) against the hindcast lead, for (a) winter, (b) spring, (c) summer and (d) autumn. The blue straight dashed line represents the statistical significance threshold at 95% confidence level. The red dashed straight line represents an arbitrary threshold for useful forecast as derived by Hollingsworth et al. (1980).
Fig. 5
Seasonal cross-correlation of the Niño3.4 Index hindcasts with the NSW model in (a) winter, (b) spring, (c) summer and (d) autumn. The values connected with a dot–dashed line were obtained with the OS predictive scheme and those connected with a dashed line with the FSM scheme. On the background we have depicted the cross-correlation of the hindcast produced assuming the Niño3.4 Index persistence with solid grey line. The blue straight dashed line represents the statistical significance threshold at 95% confidence level. The red dashed straight line represents and arbitrary threshold for useful forecast as proposed by Hollingsworth et al. (1980).
Fig. 6
Seasonal cross-RMSE of the Niño3.4 Index hindcasted with the NSW model in (a) winter, (b) spring, (c) summer and (d) autumn. The values connected with a dot–dashed line were obtained with the OS predictive scheme and those connected with a dashed line with the FSM scheme. On the background we have depicted the cross-RMSE of the hindcast produced assuming the Niño3.4 Index persistence with a solid grey line. The blue and red straight dashed lines represent the same thresholds as in Fig. 4.
Fig. 7
Seasonal cross-correlation of the Niño3.4 Index hindcasted with the NWTSA model in (a) winter, (b) spring, (c) summer and (d) autumn. The values (squares) connected with a dot–dashed line were obtained with the OS predictive scheme and those (filled square) connected with a dashed line with the FSM scheme. As in Fig. 5 the solid grey line represents the cross-correlation obtained assuming persistence. The blue and red straight dashed lines represent the same thresholds as in Fig. 5. Individual symbols (filled circle, triangle, diamond and star) represent the cross-correlation of those models (NWISV, NWPMM, NWNTZG and NWSTZG) that at a given lead score better than the ones obtained at that lead with the NWTSA model.
Fig. 8
Seasonal cross-RMSE of the Niño3.4 Index hindcasted with the NWTSA model in (a) winter, (b) spring, (c) summer and (d) autumn. The values (square) connected with a dot–dashed line were obtained with the OS predictive scheme and those (filled square) connected with a dashed line with the FSM scheme. On the background we have depicted with solid grey line, the cross-RMSE of the persistence hindcast. The blue and red straight dashed lines represent the same thresholds as in Fig. 4. Individual symbols (filled circle, triangle, diamond and star) represent the cross-RMSE of those models (NWISV, NWPMM, NWNTZG and NWSTZG) that at a given lead score better than the ones obtained at that lead with the NWTSA model.
Table 3. Acronyms and specifications of the regions used to build the extratropical predictors
NPNorth Pacific20°N–90°N140°E–120°WSPSouthern Pacific20°S–90°S160°E–80°WGSEGlobal South Extratropic20°S–90°S0°–360°RBRoss–Bellingshausen60°S–90°S160°E–60°WSPASouthern Pacific Antarctic50°S–90°S160°E–80°WSAIASouthern Atlantic and Indian Antarctic50°S–90°S60°W–150°ESAISouthern Atlantic and Indian20°S–90°S60°W–150°E
Table 2. Acronyms and variables of some tropical models used in this study
NWNIÑO3.4 and WWV IndexNSWNIÑO3.4, SOI and WWV IndexNWTSANIÑO3.4, WWV and TSA IndexNWISVNIÑO3.4, WWV and ISV IndexNWPMMNIÑO3.4, WWV and PMM IndexNWNTZGNIÑO3.4, WWV and NTZG IndexNWSTZGNIÑO3.4, WWV and STZG Index
Fig. 9
(a) The optimal growth structure obtained with NSW model for the summer Niño3.4 Index hindcast initialised from the preceding spring. (b) As in (a) but for the autumn Niño3.4 Index also initialised from spring.
Fig. 10
(a) Taylor diagram displaying a statistical comparison (cross-correlation, cross-RMSE and standard deviation) of the hindcasts for the summer Niño3.4 Index, performed with different models initialised in summer. The model variables are the Niño3.4 Index, the WWV Index and another index identified from different extratropical regions: GSE (red), NP (black), SPA (blue), RB (light blue), SAI (green) and SP (brown). (b) As in (a) but for the autumn Niño3.4 Index hindcasts with the models initialised in autumn.
Fig. 11
The figure shows the spatial pattern (EOF) associated with the third PC GSE of the MTT field: (a) for the summer and (b) for the autumn.