*3.1. Atmospheric Indices*

Of the indices presented in Table 2, the SOI has the longest history [21,22]. It is composed of the standardised pressure difference between Tahiti and Darwin. These locations are sometimes referred as "centres of action" because they are in the general region at either end of the barometric seesaw that straddles the Pacific and so demonstrate the maximum climate station-based variance in pressure

during an ENSO event. The SOI swings between positive and negative values with a phase shift from La Niña to El Niño such that when the pressure is above (below) average in Darwin and below (above) average in Tahiti, as found during an El Niño (La Niña) event, the SOI is negative (positive) (Figure 1).

**Figure 1.** Time series of Southern Oscillation Index (SOI) and Tahiti and Darwin pressure, 1997–2016. Sourced and redrawn from the Climate Diagnostics Bulletin (http://www.cpc.ncep.noaa.gov/ products/CDB).

The SOI is calculated in two stages. First, sea level pressure is standardised in relation to a set reference period, separately for Darwin and Tahiti. The differences in the standardised values between the two locations are then standardised. The resulting SOI values vary between −2.5 and 2.5, with roughly 66 percent of the values occurring between −1.0 and 1.0 [21]. Although this range of values implies symmetry around a mean of zero, there is a slight asymmetric distribution of SOI values because the strongest El Niño events tend to produce greater negative departures from zero compared to the positive departures for strong La Niña events. The strongest El Niño events are more intense than the strongest La Niña events. Although SOI values can be calculated for daily and weekly timescales, it is best if monthly to seasonal values are used in health impact analyses. This is because short term fluctuations in pressure at the two reference stations can occur due to weather and climate phenomena other than ENSO. The method of averaging over longer time scales therefore facilitates identification of continued periods of positive or negative departure of the SOI that is most likely due to ENSO.

While there are good historic reasons as to why Darwin and Tahiti were selected as the reference locations for the development of the SOI, their location is slightly south of the main equatorial region where ENSO manifests itself. Accordingly, an alternative form of the SOI was developed: the Equatorial SOI (EQ SOI) [21]. This is calculated using re-analysis as opposed to observed atmospheric pressure data, as the standardised anomaly of the difference between the area-average monthly sea level pressure between largely oceanic equatorial regions in the eastern Pacific (80◦ W–130◦ W, 5◦ N–5◦ S) and Indonesia (90◦ E–140◦ E, 5◦ N–5◦ S) (Figure 2). Although the EQ SOI may have advantages over the SOI in that it is derived for equatorial slices that more closely map onto ENSO centres of action, the record only extends back to 1949 (historical extent of the re-analysis data); the SOI is available from the late 19th century. Further in relation to the EQ SOI, it is worth mentioning that prior to the satellite era (pre-1979), the re-analyses on which the EQ SOI is based, possess some uncertainties, as in situ observations were sparse, thus compromising the quality of the re-analysis data.

**Figure 2.** Locations for calculation of SOI and Equatorial SOI. EQ is equator, DAR is Darwin, TAH is Tahiti, EPAC is eastern Pacific and INDO is Indonesia. Sourced and redrawn from Barnston [21] https://www.climate.gov/news-features/blogs/enso/why-are-there-so-many-ensoindexes-instead-just-one.

During ENSO events, the major zones of deep convection that produce thunderstorm-related rainfall move eastward away from their "normal" regions of predominance in the western Pacific. This shift is evident from rainfall observations and from space as changes in cloud patterns captured by satellite images of the global tropics. Because clouds, like all other objects, emit longwave radiation, satellite-based measurements since the late 1970s have been used to construct an Outgoing Longwave Radiation (OLR) index that has proven to be a good indicator of ENSO events (Figure 3). As described by the Stefan-Boltzman Law, the cooler an object, the lower the amount of longwave radiation emitted. Therefore, deep convective storms that reach high into the troposphere and produce substantial rainfall will have very low cloud top temperatures. Accordingly, such clouds will emit less OLR than their warmer and shallower counterparts, such that low (high) values of outgoing longwave are taken to mean enhanced (suppressed) thunderstorm activity and anomalously high (low) rainfall. Although the OLR index has yet to be used to explore ENSO health links, it may offer some potential for exploring rainfall-sensitive health outcomes, especially for the geographic region for which the index is derived (Table 2), because it is a proxy of thunderstorm/rainfall activity.

**Figure 3.** Time series of outgoing longwave radiation (see Table 2), 1997–2016, showing clearly the below average OLR for the 1997–1998 and 2015–2016 El Niño events. Sourced and redrawn from the Climate Diagnostics Bulletin http://www.cpc.ncep.noaa.gov/products/CDB.

Multiple lower and upper atmosphere wind indices have been developed for monitoring the flow of air in the lower and upper branches of the Pacific Walker Circulation [21]. The three 850 hPa indices (Table 2) represent the strength of the easterly trade winds in ENSO critical regions along the equator (Figure 4). The trade winds form the lower east to west branch of the Walker Circulation. The 200 hPa zonal wind index (Table 2) provides a measure of wind strength in the upper west to east branch of the Walker Circulation (Figure 4). At the western and eastern extremities of the Walker Circulation, air ascends and descends, respectively, thus forming the ascending and descending branches of the along the equator circulation. Positive (negative) values of the 850 hPa wind indices indicate strong

(weak) trade winds. The weakened trade winds of the 1997–1998 and 2015–2016 ENSO events are clearly visible in time series of this index for the three reference regions, especially for the 1997–1998 event (Figure 4). How inter-annual variations in the trade wind strength might play out in terms of health impacts, especially in countries directly affected by these anomalies across the wider Pacific Basin, remains to be explored.

**Figure 4.** Time series of ENSO wind indices for various regions (see Table 2) 1997–2016. The weakening of the zonal winds is especially apparent for the 1997–1998 El Niño event. Sourced and redrawn from the Climate Diagnostics Bulletin http://www.cpc.ncep.noaa.gov/products/CDB.
