Australian Rainfall Increases During Multi‐Year La Niña

Australia is one of the regions strongly affected by the El Niño‐Southern Oscillation (ENSO). The recent 2020–2023 La Niña event was marked by record‐breaking rainfall and flooding across eastern Australia. The continuous wet conditions during the triple La Niña motivated us to explore the impacts of single‐year and multi‐year ENSO events on Australian rainfall using observational data sets. We find that, while there is no difference in the rainfall impacts during single or double El Niño events, Australian rainfall tends to increase in the third year of triple La Niña events compared to the first and second years. The enhanced rainfall impact during the third La Niña year occurs despite no strengthening of La Niña in the tropical Pacific, suggesting that other processes such as local rainfall‐soil moisture feedback may play a role in prolonging the effects of multi‐year La Niña events in Australia.

These rainfall patterns, however, exhibit significant variations from event to event (Tozer et al., 2023), in part due to ENSO asymmetries and diversity in their amplitudes, spatial patterns and durations (e.g., Capotondi et al., 2015).For instance, sea-surface temperature (SST) anomalies associated with El Niño are overall of stronger magnitude than the La Niña SST cooling.In addition, La Niña often reintensifies in austral summer following the first summer peak and in some cases extends to a third year, while El Niño events usually only have one peak (e.g., Dommenget et al., 2013;Okumura & Deser, 2010).The nonlinearity of the atmospheric response to a warming in the tropical Pacific and the stochastic nature of weather systems also influence the complexity of ENSO events (e.g., Timmermann et al., 2018).
Consequently, the atmospheric teleconnections to Australia can exhibit significant variations across ENSO events and some events do not produce the expected impact (Taschetto et al., 2010;Tozer et al., 2023;van Rensch et al., 2019;Wang & Hendon, 2007).Indeed, it has been noted that El Niño SST anomalies are a poor predictor of dry conditions in Australia, while the magnitude of La Niña is closely related to the rainfall response (Power et al., 2006).
Historically, the influence of La Niña on Australian rainfall has shown to be more consistent, with wet conditions over the eastern half of the country as observed, for example, during 1998-1999 and 2010-2012.This latter event was marked by a double La Niña event and record-breaking rainfall across large parts of Australia (Ummenhofer et al., 2015).The consecutive occurrence of La Niña led to April 2010 to March 2012 being recorded as Australia's wettest 2-year period at that time (Australian Bureau of Meteorology, 2012).
The recent 2020-2023 triple La Niña event was associated with similar heavy rainfall impacts.While La Niña persisted through the 2020 austral spring and 2020-2021 summer, the associated heavy rainfall increased soil moisture, runoff and water storage levels, raising the risk of floods in many areas of eastern and southeastern Australia.Increased rainfall continued throughout the 2021 autumn, leading to flooding in many parts of southeast Australia in March of that year.This recent record-breaking rainfall raised the question of how much that extreme event was influenced by the pre-wet conditions from the first and second La Niña years, and whether there are any differences in the intensity of the rainfall response in multi-year La Niña and El Niño events compared to single-year La Niña and El Niño events.In this study, we investigate Australian rainfall impacts associated with multi-year ENSO events.Using 124 years of observational SST and precipitation data, we show that eastern Australia experiences a significant increase in rainfall during the third year of triple La Niña events.

SST Data and ENSO Identification
We use the Extended Reconstructed SST Version 5 (ERSSTv5; Huang et al., 2017) data set from January 1900 to April 2023 to calculate the Niño3.4index, following the method of the National Oceanic and Atmospheric Administration's (NOAA) Oceanic Niño Index (ONI; https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php).The ONI is calculated as the 3-month running mean monthly SST anomalies averaged over the Niño3.4region (5°N-5°S, 120°-170°W).SST anomalies are calculated relative to a 30-year base period, which is updated every 5 years to account for long-term changes in the background SST.An El Niño event is identified when the ONI exceeds a threshold of 0.5°C for at least five consecutive months of the running mean, including during the peak of the event in November and December of year zero, and the following January in year 1 (ND(0)J(1)), while a La Niña event is identified when the index is consistently below 0.5°C for those conditions.For simplicity, we omit the developing and decaying ENSO years in the month abbreviations in the text, for example, NDJ instead of ND(0)J(1).Based on this method, we identify 39 El Niño and 39 La Niña years (Figure 1a).Of these, there are 20 single, 8 double and 1 triple El Niño events, and 12 single, 6 double and 5 triple La Niña events (the years are listed in Table S1 in Supporting Information S1).The single, double and triple La Niña events match those identified by Jong et al. (2020) from 1950 to 2014, the period of their analysis.

Rainfall Data and Analysis
We use monthly rainfall data from the Australian Gridded Climate Data (AGCD; Australian Bureau of Meteorology, 2019;Evans et al., 2020), from January 1900 to February 2023.The AGCD combines rain gauges across Australia and state-of-the-art statistical analysis to provide a gridded product with a 0.05°spatial resolution, which we remap to a 0.25°grid.Soil moisture from the Australian Landscape Water Balance simulated data (Australian Bureau of Meteorology, 2023) is used over the same period as the rainfall data.Correlation analysis between the ONI and precipitation is used to determine the season that has the strongest relationship with ENSO variability and scatter plots and composite analysis is used to examine asymmetries in the rainfall response during the first, second, and third year of ENSO events.Statistical significance is assessed using a two-sided t-test.

Relationship Between ENSO and Australian Rainfall
The relationship between seasonal rainfall across Australia and the mature phase of ENSO, that is, NDJ ONI, is overall stronger during the developing and peak phases of ENSO in austral winter (June-August; JJA; Figure 1b), spring (September-November; SON; Figure 1c) and summer (December-February; DJF; Figure 1d).This is consistent with previous studies (e.g., McBride & Nicholls, 1983;Risbey et al., 2009).The strongest correlation occurs in SON, covering 70% of the country with coefficients lower than 0.3 and 30% of the country with coefficients lower than 0.4, primarily in eastern Australia.These seasonal patterns are qualitatively similar when conducting concurrent correlation analysis instead of using the peak of the ONI (Figure S1 in Supporting Information S1).
Similar to the correlation analysis, the composited rainfall response covers much of the eastern two-thirds of Australia in SON, with widespread negative anomalies during El Niño and positive anomalies during La Niña (Figures S2b and S2d in Supporting Information S1).However, the ENSO-related rainfall response is asymmetric in DJF and March-May (MAM), with stronger rainfall anomalies during La Niña than El Niño (Figure S2 in Supporting Information S1).
While El Niño events are generally stronger than La Niña events, they tend to terminate earlier as the western Pacific surface winds reverse direction with the arrival of the Australian monsoon season, counteracting the El Niño wind anomalies (Dommenget et al., 2013;Okumura & Deser, 2010).Thus, El Niño is strongly phase-locked to the seasonal cycle and lasts on average 1 year, sometimes up to 18 months (e.g., Dommenget et al., 2013;Okumura & Deser, 2010).Our classification only identifies a rare case of El Niño persisting for three consecutive austral summers during 1939-1941, however paleoclimate reconstructions of ENSO suggest that protracted El Niño events may have occurred prior to the instrumental record (Allan et al., 2020).In contrast, La Niña events typically last around 18 months, and it is common for the cooling in the tropical Pacific to extend through MAM and into the second austral summer.About 50% of the La Niña events classified in our study are multi-year events (Table S1 in Supporting Information S1).The SST cooling in the tropical Pacific can sometimes last for 3 years,  1)) Niño3.4 index and September-November (SON(0)) mean precipitation averaged over the east Australian region (east of 138°E, north of 39°S) where the SON(0) correlation coefficient is below 0.4 (as in Figure 1c).Green crosses represent single ENSO events, blue-tone squares show double ENSO events, and red-tone triangles highlight triple ENSO events.producing a triple La Niña.About 22% of the La Niña events classified in our study are triple events (Table S1 in Supporting Information S1), consistent with the finding by Jong et al. (2020).

Geophysical Research Letters
To illustrate the differences in the magnitude of the rainfall anomalies between El Niño and La Niña, we create a time series of rainfall anomalies over the region with the strongest relationship to ENSO, that is, the east Australian region (east of 138°E, north of 39°S) where the SON correlation coefficient is below 0.4 (as in Figure 1c).Figure 2 shows a scatter plot between the NDJ ONI and SON rainfall averaged over that region.The relationship between ENSO and east Australian rainfall is clearly nonlinear.The correlation coefficient for the negative ONI values is 0.46, which is statistically significant at the 99% level, while the correlation for the positive ONI values is 0.28 (p < 0.05).This means that La Niña events generally promote a larger precipitation response than El Niño.This nonlinear relationship has been reported in previous studies (Chung & Power, 2017;King et al., 2013;Power et al., 2006).Here we further show that this nonlinear relationship persists over multiyear ENSO events.
An interesting feature visible in Figure 2 is that the later years of multi-year La Niña can be accompanied by high rainfall anomalies (see red triangles) even for weak ONI magnitude.For triple La Niña, the positive rainfall anomaly tends to increase as the multi-year event progresses from the first year to the third year, with the highest anomaly in the third SON and a large increase in the rainfall anomaly from the second to the third SON in four of the five identified triple La Niña events, including the recent 2020-2023 event.The exception to this rule is the 1973-1975 triple La Niña event when the highest rainfall occurred in the first year and the lowest rainfall in the second year.Furthermore, the relationship between the magnitude of the ONI and the associated rainfall anomaly for triple La Niña events can be the opposite to the usual relationship for all La Niña events: the wettest SON in a triple La Niña occurs in the year with the weakest ONI value in three of the five triple La Niña events.
We further illustrate in Figure 3 the evolution of precipitation and the ONI in the first and second years of El Niño events (Figures 3a and 3c), and in the first, second and third years of La Niña events (Figures 3b-3d).Despite the large spread across ENSO-associated rainfall events, there are overall negative precipitation anomalies during El Niño (Figure 3a) and positive anomalies during La Niña (Figure 3b), with the largest anomalies occurring near December.In addition, there exists an increase in the peak rainfall anomaly over time for multi-year ENSO.For both El Niño and La Niña, the peak median second-year precipitation anomaly centered on December ( 12.33 mm for El Niño and 14.51 mm for La Niña) is greater than the first-year value ( 12.29 mm for El Niño and 10.66 mm for La Niña) of all events.This increase is more obvious for La Niña events, where the median precipitation anomaly peak increases further to 37.76 mm in the third year, indicating approximately 2.5 times more rainfall from the first La Niña December to the third La Niña December.The median precipitation calculated separately for double and triple events also supports the increased rainfall in the third year of a multiyear La Niña event (not shown).This suggests that multi-year La Niña events have stronger impacts on eastern Australian rainfall over time.A similar strengthening of the multi-year La Niña teleconnection to USA wintertime precipitation has been observed by Okumura et al. (2017).
It seems a counter-intuitive finding that the rainfall impacts are stronger in the later years of a multi-year ENSO event since the ONI SST anomalies are not necessarily stronger in the second and third years of multi-year events (Figures 3c and 3d).Both El Niño and La Niña events typically peak around December, except for the last year of multi-year events which tends to develop slightly earlier: the second-year El Niño peaks in October (Figure 3c) and the third-year La Niña strengthens from June, peaks in October, and quickly decays after December (Figure 3d).Unlike the east Australian precipitation anomalies, the median ONI peaks do not tend to increase in the second and third years compared to the first years in all events.The SST anomalies show a similar spread near each peak for both El Niño and La Niña events, indicating no clear difference in SST strength in the second or third years.Iwakiri and Watanabe (2022) argue that the evolution of multi-year El Niño and La Niña events is roughly symmetric, except for multi-year La Niña following strong El Niño events.Those multi-year La Niña events generally have a stronger first year than subsequent years (e.g., the 1973-1975La Niña following the 1972 El Niño).This is because strong El Niño events promote intense discharge of warm subsurface water from the equatorial Pacific and strong thermocline feedback (e.g., Dommenget et al., 2013), favoring a sharp and intense transition to a tropical Pacific cooling.However, multi-year La Niña events that are not preceded by a strong El Niño show a time evolution that is of similar strength to multi-year El Niño events, with no significant difference in strength between the first, second and third years.Therefore, despite no significant strengthening of ONI in the second or third year of multi-year La Niña, the east Australian rainfall response during the monsoon peak intensifies in the third year of multi-year La Niña events.There were only five triple La Niña events in the past 124 years, leading to a small sample size that could cause uncertainty when examining differences in rainfall between years.Therefore, we investigate the statistical significance of this result using a bootstrapping test where the rainfall anomaly around December (when it peaks) is compared with the anomalous rainfall during multi-year ENSO events.The bootstrapping test is conducted between the NDJ mean precipitation of all years and NDJ mean precipitation of each of the first and second El Niño years and the first, second and third La Niña years.Except for the second-year El Niño NDJ, the other years all have a statistically significant difference in mean precipitation at the 95% confidence level based on 1,000 samples randomly selected with replacement from the observations.In addition, a bootstrapping test is conducted between the NDJ mean precipitation of the first-year La Niña, second-year La Niña, third-year La Niña, first-year El Niño, and second-year El Niño events.The differences in NDJ mean precipitation between the 3 years of triple La Niña events are only statistically significant when the first or second year is compared to the third year (firstyear La Niña precipitation compared to third year, and second-year La Niña precipitation compared to third year).In other words, the NDJ precipitation in the third year of triple La Niña events is significantly different from that in the first and second years.The statistical significance of this result has also been confirmed using a Student's ttest and a non-parametric Mann-Whitney U statistic at the 95% level (Table S2 in Supporting Information S1).
We further examine spatial composites of 3-month mean precipitation anomalies in the first, second and third years of El Niño and La Niña events (Figure 4).The composite for individual months is displayed in Figure S3 in Supporting Information S1.SON shows a larger area of Australia with consistent rainfall responses during multi-year ENSO, while sparse anomalies in the north and east appear during the summer months.This seasonality is perhaps not surprising given the Australian tropics receive their largest proportion of annual rainfall during summer when the monsoon is well-established.As the multi-year event continues, the anomalous rainfall patterns intensify in certain areas for El Niño and La Niña.Specifically, the dry anomalies associated with El Niño are generally significant in the first year (Figures 4a-4d) and remain significant in the second year over northeastern Australia (Figures 4e-4h).The wet anomalies during multi-year La Niña are of stronger magnitude than the El Niño-related drying, possibly due to the skewed nature of rainfall distribution.However, the statistical significance of the La Niña rainfall response is reduced compared to El Niño due to the relatively smaller sample size and larger internal variability (e.g., Figures 3a and 3b).Nevertheless, the increase in rainfall surplus and area extent during the third year of multi-year La Niña events is prominent in Figure 4.

Discussion and Conclusions
We examined the changes in Australian rainfall anomalies during multi-year ENSO events.Our findings show that the east Australian rainfall response to multi-year La Niña increases from the first to the third year of the event.This suggests that the impact of La Niña on rainfall in eastern Australia becomes stronger over time.This rainfall pattern was observed during the most recent triple La Niña in 2020-2023, as well as previous triple La Niña events, when the spring-to-summer rainfall in the third year was higher than in the preceding 2 years.In contrast, the mild strengthening of the east Australian rainfall response in the second year of double El Niño events is overall not statistically significant.
The increase in rainfall during multi-year La Niña events occurs despite the ONI in the second and third years of multi-year La Niña events being weaker or of similar strength compared to the first year's value.While several factors could play a role in modulating this rainfall increase, such as local SST (e.g., van Rensch et al., 2019), a combination of negative IOD and positive SAM (e.g., Hendon et al., 2014;Ummenhofer et al., 2011), or multidecadal modulation by the Interdecadal Pacific Oscillation (e.g., Heidemann et al., 2022;Power et al., 1999), a more plausible mechanism to explain this rainfall increase could arise from a soil moisturerainfall feedback.
Cumulative hydrological effects integrated into deep layer soil moisture can provide long-term memory for rainfall beyond a season and even at multi-year timescales (Sharmila & Hendon, 2020).Prolonged rainfall during La Niña leads to high/saturated soil moisture (Figure S4 in Supporting Information S1), which, combined with increased evapotranspiration and water recycling (Holgate et al., 2020(Holgate et al., , 2022)), can generate an enhanced rainfall response in austral summer.Although the Pacific SST anomalies weaken in the summer of the third year of a triple La Niña event (Figure 3d), La Niña develops earlier in austral winter and terminates earlier than the first and second years (Figure 3d and Figures S5 and S6 in Supporting Information S1).This leads to an earlier and intensified rainfall response in austral spring (Figure 4), when the ENSO-Australian rainfall relationship is strongest (Figure 1), thus increasing soil moisture prior to the monsoon season (Figure S4 in Supporting Information S1).The increased soil moisture intensifies rainfall recycling during the monsoon season, leading to increased rainfall response in the summer of the third La Niña year.
In addition to the soil moisture-rainfall feedback, a more complex self-sustaining mechanism involving atmosphere-ocean-land interactions may operate in the third La Niña year.Sekizawa et al. (2018) showed that multi-year variations in northern Australia wet season (November-April) rainfall are partly driven by a rainfallwind-evaporation and local soil moisture-rainfall feedback.During summer, enhanced convection and rainfall over northern Australia strengthen the cyclonic circulation associated with the monsoonal low, intensifying the surface westerlies over the southeast Indian Ocean, off the northwest of Australia.These anomalous westerlies increase wind speed, promoting evaporation over the warm ocean and advecting moisture inland.This leads to increased precipitation and increased soil moisture over land, providing memory and enhancing evapotranspiration in the succeeding years.The high soil moisture combined with increased evapotranspiration generates increased rainfall in the monsoon season of the third La Niña year.This self-sustained mechanism occurs primarily over northern Australia, even when ENSO is neutral (Sekizawa et al., 2023).Therefore, it is possible that when the Pacific SST weakens and terminates earlier than in the previous 2 years (Figure S5 in Supporting Information S1), the rainfall-wind-evaporation feedback acts to intensify rainfall in the third year of the triple La Niña.By extending the ONI back to 1900, we have identified a triple El Niño event between 1939 and 1941, and five triple La Niña events since 1900.Although the sample size is limited and the reliability of historical data may be questioned, this extended classification has revealed differences in the characteristics of the Australian rainfall response between single-year and multi-year ENSO events.Future work may investigate these findings using large-ensemble climate model simulations to increase the sample size, while recognizing that climate models have limitations in representing the Australian monsoon, ENSO dynamics, and associated teleconnections (Jourdain et al., 2013;Taschetto et al., 2014).
Understanding the impacts of multi-year ENSO teleconnections on Australian rainfall is crucial for improving seasonal forecasting and informing stakeholders and decision-makers on rainfall predictions and extremes.Our findings provide an important message for managing timely responses to possible heavy rainfall and increased flood risk during future triple La Niña events.This work was possible thanks to the support from the Australian Research Council (ARC) Centre of Excellence for Climate Extremes (CE170100023).A.S.T. acknowledges support from the Australian Government's National Environmental Science Program.The authors thank two anonymous reviewers for their valuable feedback on the manuscript.Open access publishing facilitated by University of New South Wales, as part of the Wiley -University of New South Wales agreement via the Council of Australian University Librarians.

Figure 1 .
Figure 1.El Niño-Southern Oscillation time series and its relationship with Australian rainfall.(a) Time series of the Oceanic Niño Index (ONI).A threshold of ±0.5°C (dashed red and blue lines), is used to identify El Niño (red shading) and La Niña (blue shading) events.Correlation coefficients between the ND(0)J(1) ONI and seasonal precipitation anomalies in (b) JJA (0), (c) SON(0), (d) D(0)JF(1), and (e) MAM(1).Contour intervals of 0.2 are represented by dashed lines for negative correlation coefficients.The percentage of Australian area with correlations lower than 0.3 and 0.4 is noted at the bottom right of each panel.Stippling shows correlation coefficients that are statistically significant at the 95% level based on a twosided t-test.State/territory boundaries are highlighted in (e): WA = Western Australia, NT = Northern Territory, SA = South Australia, QLD = Queensland, NSW = New South Wales, ACT = Australian Capital Territory, VIC = Victoria and TAS = Tasmania.

Figure 2 .
Figure 2. Asymmetry in El Niño-Southern Oscillation teleconnections to Australian rainfall.Scatter plot showing the relationship between the November-January (ND(0)J(1)) Niño3.4 index and September-November (SON(0)) mean precipitation averaged over the east Australian region (east of 138°E, north of 39°S) where the SON(0) correlation coefficient is below 0.4 (as in Figure1c).Green crosses represent single ENSO events, blue-tone squares show double ENSO events, and red-tone triangles highlight triple ENSO events.

Figure 3 .
Figure3.Evolution of El Niño-Southern Oscillation and east Australian rainfall.Timeseries of precipitation anomalies for all (a) El Niño and (b) La Niña events between January 1900 and February 2023.Oceanic Niño Index for all (c) El Niño and (d) La Niña events between January 1900 and April 2023.The monthly precipitation anomaly is the average precipitation anomaly over the east Australian region (east of 138°E, north of 39°S) where the correlation coefficient with the Niño3.4index is smaller than 0.4.A 5-month running mean is applied to the timeseries.The shaded region shows the positive and negative one standard deviation of the monthly precipitation or Niño3.4 index, and the black line shows the median of all events.

Figure 4 .
Figure 4. Rainfall Anomalies in El Niño-Southern Oscillation years.Composites of Australian 3-month mean rainfall anomalies (mm month 1 ) for all (a-h) El Niño and (i-t) La Niña events, distinguishing into the first, second, and third years of the corresponding event.Stippling shows areas statistically significant at the 95% level based on a two-tailed t-test.The number of samples (n) in each composite is listed in the bottom left corner of (d), (h), (l), (p) and (t).Month-by-month rainfall anomaly composites are shown in Figure S3 in Supporting Information S1.
Wet conditions continued in spring 2021 through to autumn 2022, producing Australia's wettest November in 122 years (Australian Bureau of Meteorology, 2022b) and recordbreaking floods in southeast Queensland and eastern New South Wales in late February and March 2022 (Australian Bureau of Meteorology, 2022c).Heavy rainfall continued in spring 2022, with highest on record rainfall across much of New South Wales and Victoria (Australian Bureau of Meteorology, 2022a).