Introduction:

The relative importance of observing sea surface salinity (SSS) to improve statistical predictions of El Niño/Southern Oscillation (ENSO) is examined here. This study focuses on the use of linear multiple regression to predict the time evolution of the NIÑO3 index (the average of SST over the region [150W-90W,5N-5S].
Five series of multiple regression models are constructed between the the NINO3 index and lagged time series of observed sea surface temperature (SST), SSS, sea surface height (SSH), Precipitation minus evaporation (P-E), zonal wind stress (STX) and meridional wind stress (STY). Time lags are 0, 3, 6, 9, and 12 months. The forward stepwise method is used to select the relevant modes. Partial F-tests are used to accept a new variable. A variable selected at one moment may be rejected after the introduction of another variable. The presence or not of SSS predictors in the resulting statistical model indicates if and where SSS will provide significant new information.

Data:

SST fields come from the analysis of Reynolds and Smith [1994]. Wind stress is derived from FSU pseudo-wind stress [Stricherz et al., 1992]. SSS fields come from the gridded data of Delcroix [1998]. SSH fields are obtained from the Climate Modeling Branch (CMB) ocean hindcasts of the NCEP Environmental Modeling Center [Behringer et al., 1998]. Precipitation is obtained from the merged analysis using rain gauge and satellite estimates [ Xie and Arkin, 1996]. Latent heat fluxes come from the Climate Data Assimilation System I (CDAS-I) reanalysis of NCAR [Kalnay et al., 1996]. All fields are interpolated over a common 10o-longitude by 2o-latitude grid of the tropical Pacific Ocean [140E-80W,20S-20N]. Anomalies are calculated around the monthly climatology calculated over 1980-1995. The time period spans from January 1980 to December 1995, i.e. the period for which all data fields are simultaneously available. Figures 1A and 1B show the longitude-time diagrams averaged around the equator (2S-2N). The units are oC for temperature, psu for salinity, cm for sea level, mm/month for P-E, and dyn/cm2 for wind stress.

Figure 1. Longitude time diagrams of the anomalies of SST, SSS, SSH, Precipitation minus evaporation (P-E), zonal wind stress (STX) and meridional wind stress (STY). Time period: January 1980 to December 1995. Units are C for temperature, psu for salinity, cm for sea level., mm/month for P-E, and dyn/cm2 for wind stress.

Results:

Figure 2 shows the correlation (lag zero) between the NIÑO3 index and the data fields. Shaded areas indicate non-significant correlation. The largest SSS correlation is located at the western part of the equatorial region, and are related to the eastward displacement of the local rainfall. The stepwise method identifies the relevant, non-redundant predictors. Sensitivity analysis are used to discard non-robust predictors. At zero lag, only four predictors are retained. They are indicated by letters A, B, C, and D (Figure 2). All those predictors correspond to SST in agreement with the idea that the most significant variable to nowcast the NIÑO3 index is the SST itself. The statistical models constructed for 3 months predictions contain only one predictor, an SSH point (not shown). The SSS correlation map is similar to the one shown in Figure 2 with a region of strong positive correlation (0.6) at [150W,15S].

Figure 2: Correlation between SST, SSS, SSH, P-E, STX, and STY with the Niño-3 index. Yellowed regions indicate nonsignificant correlation (90% confidence). It is assumed that a new degree of freedom (DOF) is acquired every 6 months. N=192 months, DOF=32.

6-month (Figure 3), 9-month, and 12-month (Figure 4) prediction models select SSS predictors. The amount of NIÑO3 variance explained by the SSS-predictors is 36%, 6% and 25% respectively. For short lags the region of largest correlations is located at the western edge of the equator. For longer lead times, the region of maximum correlation is located South of the equator and in the central part of the domain. From such a region, SSS anomalies have the potential to modify the subsurface stratification of the western Pacific via subduction of positive anomalies as they propagate westward. Figure 5 shows the composites of SSS anomalies 12 months before a warm (top) and cold (bottom) events.

Figure 3: As Figure 2, but fields lead by 6 months. N=186, DOF=31

Figure 4: As Figure 2, but fields lead by 12 months. N=180, DOF=30

Conclusions:

We have shown that SSS variability in the western tropical Pacific (WTP)  is significantly correlated with the NIÑO3 index and that SSS predictors provide independent information with respect to SST and SSH. We claim that SSS monitoring has a potentially important, non redundant role in 6-12 month forecasts. For these lags, regions with high correlation between SSS and NIÑO3 index are found off the equator, in areas of subduction and susceptible to tropical/extra-tropical interactions.
The general lack of SSS observations, particularly off the equator is limiting our understanding of the role of SSS anomalies in the preconditioning of ENSO and the decadal modulation of the equatorial heat content. For example, the El Niño-La Niña event of 1997-98 did produce a sea level anomaly in the south WTP several months before the peak of the warm event. Strong upwelling associated with this anomaly would result in positive anomalies consistent with Figure 5, which have a negative impact on sea level. Estimation of the thermal content is under-estimated without consideration of SSS variability. Similarly, negative sea level anomalies are found south of the equator more than one year before the peak of La Niña of 1998. If the relationship of Figure 4 is valid, negative SSS anomalies might have existed in the WTP. Estimation of the heat content without consideration of the SSS variability over-estimates the heat content of the WTP prior to the 1998 La Niña.

Reference:

Ballabrera-Poy, J., R. Murtugudde, and A.J. Busalacchi, On the potential role of sea surface salinity observations on ENSO predictions, J. Geophys. Res., 2002.

HOME| LINKS | NEWSROOM | SITE MAP OVERVIEW | SCIENCE | TECHNOLOGY & OPERATIONS | GALLERY | EDUCATION