**4. ENSO Indices for Climate and Health Analyses**

Faced with a range of ENSO indices, questions that are likely to arise when planning analyses of ENSO health associations are "which index will be optimal for exploring ENSO-related health impacts?" and "are there different ways of defining ENSO events?"

Currently there is no common consensus in the climate science community as to which ENSO index best describes ENSO phases. This appears to contrast with the climate and health science community in which there appears to be a substantial amount of blind faith applied to the use of ENSO indices for exploring possible ENSO driven health variations at a range of temporal and spatial scales. In many ways, the choice of an ENSO index for health analyses may depend on geographical location, its relation to a health-sensitive climate variable or even disease outcome. For example, for regions in close proximity to the Niño oceanic regions, the SST-based indices may be appropriate, especially for exploring rainfall, air and sea temperature-related health outcomes. Similarly, the OLR index may be useful for exploring rainfall-related health issues in the central Pacific. ENSO-related wind indices, because they represent the surface trade wind strength, are more than likely to be only useful in the trade wind regions of the Pacific Basin. For locations distant from the Pacific Basin, such as southern and eastern Asia, one of the pressure-based indices might be more suitable, as these represent variations in atmospheric circulation conditions over a larger geographical range.

While some researchers might be tempted to use one of the more recently developed blended or multivariate ENSO indices, it is worth bearing in mind that these are complex indices made up of multiple interacting variables. Accordingly, they may not be appropriate if the purpose of an analysis was to uncover a direct link between a specific climate attribute such as temperature and a disease outcome. Further, blended indices have not yet been widely adopted by ENSO forecasting centres. This is most likely because of the prediction error associated with the individual input variables such that the cumulative error for a predicted value of a blended index could be large when compared to a single variable based index. Notwithstanding this, blended indices such as that of Wolter and Timlin [38] have been applied on a number of occasions in ENSO-health studies (see Section 6).

In selecting ENSO indices, researchers also need to be aware of the producing agency because of the way ENSO phases and events are identified can vary between agencies. As noted above, NOAA uses the Niño 3.4 region (5◦ N–5◦ S, 170◦ W–120◦ W) based on ONI and the persistence of SST anomalies in excess of +/−0.5 ◦C for five months for identifying ENSO phases. In contrast, the Japan Meteorological Agency (JMA) uses a slightly different formulation for the calculation of their ENSO index, which is a 5-month running mean of spatially averaged SST anomalies over the geographical range of 4◦ S–4◦ N, 150◦ W–90◦ W; this is essentially a latitude-restricted Niño 3 region. Further, the JMA define an ENSO year as October through to the following September. If the JMA index values exceed +0.5 ◦C or −0.5 ◦C for six consecutive 5-month periods, including October to December, the ENSO year is declared as either El Niño or La Niña. The Australian Bureau of Meteorology (BoM), like NOAA and the JMA, uses SST anomalies as a basis for the ENSO phase definition but sets a higher threshold. For an El Niño (La Niña) event to be called, SST anomalies in the Niño 3 and Niño 3.4 regions must exceed +0.8 ◦C (−0.8 ◦C). Further to the SST criteria, BoM specifies much weaker (stronger) trade winds over the western or central equatorial Pacific for the previous 3–4 months, as well as a SOI value of −7 (+7) to be necessary for an El Niño (La Niña) event to be specified. Note that in the case of BoM, the SOI values are quite different from those associated with the NOAA SOI index because the BoM uses the Troup SOI, the standardised anomaly of the mean sea level pressure difference between Tahiti and Darwin. The BoM calculation uses the period 1933 to 1992 as the climatology; this contrasts with the NOAA and JMA. Once the Tromp SOI is calculated, it is multiplied by 10. Using this convention, the BoM SOI takes on values in the range of −35 (strong El Niño) to about +35 (strong La Niña).

NOAA, JMA, and BoM use SST anomalies in their definitions of the El Niño and La Niña phases of ENSO. In addition to the slight differences in the criteria used for defining events (e.g., SST anomaly, anomaly period, and region), a further source of difference between the oceanic indices are the SST data sets employed for constructing the SST anomalies. For example, the JMA uses the Centennial In Situ Observation-Based Estimates (COBE)-SST data set for ENSO monitoring with sliding climatological values based on the most recent 30-year period as described by Ishii et al. [43] and JMA [44]. In contrast, NOAA and BoM use the Extended Reconstructed Sea Surface Temperature, Version 5 (ERSSTv5) data set with anomalies based on centred 30-year periods updated every five years as described by Huang et al. [45]. Although not described here, the United Kingdom's Met Office's Hadley Centre applies yet another SST data set for deriving historical SST-based ENSO measures, the HadISST data set [46]. Important to note in the consideration of possible pre-1950 ENSO health associations is that for this period, because of observational uncertainties, the ERSSTv5 and HadISST, from which ENSO indices are derived, demonstrate significant differences [47]. Given this, researchers need to be aware of the differences in SST data sets for deriving ENSO indices in terms of the SST observations drawn upon, statistical and data assimilation methods applied, and spatial resolution of the final SST products [48] because these data set properties may affect the degree to which meaningful associations between ENSO and health outcomes can be quantified.

Hanley et al. [25] provided a useful comparison of ENSO indices in terms of their ability to describe ENSO events. They found that El Niño only engenders a weak SST response in the Niño 4 region, whereas La Niña produces quite a strong SST signal in the Niño 1 + 2 region (see Figure 6). They also concluded, based on an analysis of the sensitivity of a range of ENSO indices relative to each other, that the choice of an index for analysing ENSO related risks is somewhat dependent on the ENSO phase. For example, in the case of La Niña events, the JMA ENSO Index was more sensitive

than other atmospheric and oceanic indices. In contrast for El Niño events, the SOI, Niño 3.4, and Niño 4 indices are almost equally sensitive but more sensitive than the JMA, Niño 1 + 2, and Niño 3 indices.

ENSO indices have also been used to classify the strength of ENSO phases. Such classifications may be of interest to climate and health researchers because El Niño or La Niña strength may be an indicator of the potential scale of climate-related health risks, all other variables, such as socio-economic conditions or vulnerability, being equal. A recent classification of ENSO phase strength for the period 1950–2016 was produced by Santoso et al. [47] using Niño 3.4 SST anomalies averaged across four SST reanalysis products (ERSSTv4, ERSSTv5, HadISST, and COBE) over the months of November-December-January (NDJ) and December–January–February (DJF). Because the classification is based on SST anomalies for NDJ and DJF, it identifies the year of the development phase of an ENSO event. For strong (weak) events, the averaged Niño 3.4 anomaly must exceed 1 (be between 0.5 and 1) standard deviation. A neutral phase is deemed to be associated with a standard deviation of less than 0.5. Strong and weak El Niño and La Niña years are listed in Table 3. The classification, while identifying the often cited extreme 1972/1973, 1982/1983, 1997/1998, and 2015/2016 El Niño events as extremes, also highlights other strong El Niño events that have received little attention in the literature. Recalling that two broad types of El Niño events occur, Santoso et al. [47] used the Niño4 index DJF average to identify Central Pacific El Niño events; the Niño 4 average must be greater than 0.5 ◦C and greater than Niño3 to be classified as a CP event (Table 3). Usefully, Santoso et al. [47] also analyse the sensitivity of El Niño and La Niña strength classification to varying criteria; this has implications for general qualitative statements about ENSO strength and health associations.

**Table 3.** Years in which strong and weak ENSO phases developed (Source: Santoso et al. [48]).


Perhaps even more apposite when considering the application of ENSO indices to the analysis of ENSO-health associations would be the reflection on the conceptual links between ENSO and health-sensitive climate fields given that the ENSO signature may vary considerably with season and location.
