A New Picture of the Global Impacts of El Nino-Southern Oscillation

The El Nino-Southern Oscillation (ENSO) is the dominant interannual variability of Earth’s climate system and plays a central role in global climate prediction. Outlooks of ENSO and its impacts often follow a two-tier approach: predicting ENSO sea surface temperature anomaly in tropical Pacific and then predicting its global impacts. However, the current picture of ENSO global impacts widely used by forecasting centers and atmospheric science textbooks came from two earliest surface station datasets complied 30 years ago, and focused on the extreme phases rather than the whole ENSO lifecycle. Here, we demonstrate a new picture of the global impacts of ENSO throughout its whole lifecycle based on the rich latest satellite, in situ and reanalysis datasets. ENSO impacts are much wider than previously thought. There are significant impacts unknown in the previous picture over Europe, Africa, Asia and North America. The so-called “neutral years” are not neutral, but are associated with strong sea surface temperature anomalies in global oceans outside the tropical Pacific, and significant anomalies of land surface air temperature and precipitation over all the continents.

www.nature.com/scientificreports www.nature.com/scientificreports/ and extratropical Pacific Ocean. It is very important to note that the SST anomalies during the cold-to-warm transition (Fig. 1c) have opposite signs to those during warm-to-cold transition (Fig. 1g). Therefore, they should not be added together into a "neutral phase" composite as in many previous ENSO studies, because the significant anomalies with opposite signs will be cancelled out. Figure 1 suggests that we should use a new four-phase paradigm of ENSO (warm phase, cold phase, cold-to-warm transition, and warm-to-cold transition) to replace the traditional three-phase paradigm (warm phase, cold phase, and neutral phase).
Supplementary Fig. 2 shows a summary of the impacts of ENSO lifecycle on global SST. Different datasets demonstrate consistent significant anomalies over global oceans throughout ENSO lifecycle. Various mechanisms have been proposed for the global SST anomalies associated with ENSO, such as the "atmospheric bridge" mechanism [24][25][26] , and for persistence of SST anomaly into next year, such as the "re-emergence mechanism" 27,28 . Dynamically, the observed significant SST anomalies throughout ENSO lifecycle drive temperature, precipitation and atmospheric circulation anomalies over the globe [29][30][31][32][33][34][35][36] .   8 . During the transition phase from El Nino to La Nina (Fig. 2g), there are significant warm surface air temperature anomalies over west Africa, south Africa, south Asia, north Australia, northeast United States, and northeast Brazil, but cold temperature anomalies over Argentina. At lag + 1.5 years (Fig. 2h), cold temperature anomalies start to occupy South America, Alaska, western Canada and maritime continent. Evolution after La Nina ( Fig. 2a-d) are simply reversed in sign. These results are generally confirmed by ERA-Interim reanalysis 2-meter temperature ( Supplementary Fig. 3) and University of Delaware surface air temperature data (not shown). Overall, the summary plots ( Supplementary  Fig. 4) show significant impacts of ENSO lifecycle on land surface air temperature over all the continents, which www.nature.com/scientificreports www.nature.com/scientificreports/ are much wider than in the current schematic (Trenberth et al.) 8 . The corresponding maps of the linear regression coefficient between surface air temperature anomaly and Nino 3.4 SST anomaly ( Supplementary Fig. 5) shows that the largest temperature responses are in NH high latitudes including Canada, Alaska and Russia.  (Fig. 3g), significant wet anomalies start to develop over south India, maritime continent, west Africa and west Australia, while dry anomalies appear over Middle East. These anomalies grow much wider 6 months later (Fig. 3h). Over other continents such as north America and south America, strong ENSO impacts cover a larger area than what we knew before. For example, the significant ENSO impact on California droughts (Fig. 3a,b,e,f) is not well-depicted www.nature.com/scientificreports www.nature.com/scientificreports/ in the widely-used schematic. The results from GPCC dataset are generally supported by CRUTS dataset (Supplementary Fig. 6) and merged satellite-gauge GPCP dataset ( Supplementary Fig. 7). The corresponding maps of the linear regression coefficient between surface precipitation anomaly and Nino 3.4 SST anomaly ( Supplementary Fig. 8) shows that the largest precipitation responses are in the tropics and subtropics.
ENSO has a strong seasonality with the amplitude of Nino3.4 SST anomaly being strongest in November-December-January 37,38 . The surface air temperature anomalies in the summer after El Nino (Fig. 2f) are significantly different from those in the summer before El Nino (Fig. 2d). The precipitation anomalies in the summer after El Nino (Fig. 3f) also have important differences from those in the summer before El Nino (Fig. 3d). Studies using high Nino3.4 SST anomaly during summer tend to combine the two summers together. Our results suggest that it is important to separate the summer after El Nino from that before El Nino since they are at different stages of the ENSO lifecycle. Figure 4 demonstrates our new schematic of global impacts of ENSO lifecycle on surface air temperature and precipitation. We only include the significant temperature and precipitation anomalies that are supported by at least two different datasets. Comparison of Fig. 4 with the current schematic (Trenberth et al.) 8 shows that the impacts of ENSO are much wider than those shown in the widely-used schematic. The largest differences are in Europe, Asia and Africa. ENSO significantly modulates surface air temperature in northern Europe, north Asia, south Asia, and entire Africa, and precipitation in southern Europe, central Asia, northern north Asia, and central Africa. For other continents such as north America and south America, significant ENSO impacts cover a larger area than those shown in the widely-used schematic. It is important to note that the so-called "neutral phases" are associated with significant temperature and precipitation anomalies, and anomalies during cold-to-warm transition and warm-to-cold transition have opposite signs.

Methods
Datasets used in this study are listed in Supplementary Table 1. The ENSO index used in this study is Nino3.4 SST from ERSST dataset. We have tested Nino3.4 SST from COBE2 SST and HadISST datasets and the results are similar. We follow the traditional methods for analysing ENSO's global impacts [16][17][18][19][20][21][22] . Linear trend (a single trend calculated for all months together) and composite seasonal cycle are first removed from all datasets. To isolate the interannual ENSO signals from the decadal variability and higher-frequency variability, the anomalies are then filtered with a 3-6 year butterworth filter (Murakami) 39 . We also tested a wider 2-10 year filter and the results are similar. Lag-correlation is calculated with the Nino3.4 SST anomaly. Statistical significance is evaluated following Oort and Yienger 40 .

Data availability
Datasets used in this study are from NOAA ESRL/PSD Climate Data Archive https://www.esrl.noaa.gov/psd/ and NCAR Research Data Archive https://rda.ucar.edu.