Nonlinearities in the Evolutional Distinctions Between El Niño and La Niña Types

Abstract

Using the HadISST, SODA reanalysis, and various other observed and reanalyzed data sets for the period 1950–2010, we explore nonlinearities in the subsurface evolutional distinctions between El Niño types and La Niña types from a few seasons before the onset. Cluster analysis carried out over both summer and winter suggests that while the warm‐phased events of both types are distinguishable, several cold phased events are clustered together. Further, we apply a joint Self‐Organizing Map (SOM) analysis using the monthly sea surface temperature anomaly (SSTA) and thermocline‐depth anomalies in tropical Pacific (TP). Results reveal that the evolutionary paths of El Niño Modoki (EM) and El Niño (EL) are, broadly, different. Subsurface temperature composites of EL and EM show different onset characteristics. During an EL, warm anomaly in the west spreads eastward along the thermocline and reaches the surface in the east in March–May of year(0). During an EM, warm anomaly already exists in the central tropical Pacific and then reaches the surface in the east in September–November of year(0). Composited SSTAs during La Niña (LN) and La Niña Modoki (LM) are distinguishable only at 80% confidence level, but the composited subsurface temperature anomalies show differences in the location of the coldest anomaly as well as evolution at 90% confidence level. Thus, the El Niño flavor distinction is potentially predictable at longer leads.

Predicting clustered weather patterns: A test case for applications of convolutional neural networks to spatio-temporal climate data

“…The performance of CNN in re-identifying (for day 0) which cluster index a pattern belongs to, or predicting which cluster index a given pattern will evolve to in a few days, can be used to evaluate and explore improvements to the CNN algorithms for each specific dataset. Note that here we use K-means clustering for indexing, but other algorithms such as hierarchical, expectation-maximization, or self-organizing maps 3,4,[23][24][25][26][27][28][29] can be used instead. However, the K-means algorithm, which clusters the data into a priori specified n classes based on Euclidean distances, provides an effective, simple method for the objective here, which is labeling the dataset for evaluating CNN, as opposed to finding the most meaningful (if even possible 17 ) number of clusters in the spatio-temporal data.…”

“…Classifying, identifying, and predicting specific patterns or key features in spatio-temporal climate and environmental data are of great interest for various purposes such as finding circulation regimes and teleconnection patterns [1][2][3][4][5] , identifying extreme-causing weather patterns [6][7][8][9][10][11][12] , studying the effects of climate change [13][14][15][16] , understanding ocean-atmosphere interaction 8,17,18 , weather forecasting 8,12,19,20 , and investigating air pollution transport 21,22 , just to name a few. Such classifications/identifications and predictions are often performed by employing empirical orthogonal function (EOF) analysis, clustering algorithms (e.g., K-means, hierarchical, self-organizing maps 1,3,[23][24][25][26][27][28][29] ), linear regression, or specifically designed indices, such as those used to identify atmospheric blocking events. Each approach suffers from some major shortcomings (see the reviews by Grotjahn et al 6 and Monahan et al 30 ); for example, there are dozens of blocking indices which frequently disagree and produce conflicting statistics on how these high-impact extreme-causing weather patterns will change with climate change 10,14,31 .…”

Why Does the CP El Niño less Frequently Evolve Into La Niña than the EP El Niño?

“…Kao and Yu (2009) noticed that the EP El Niño and La Niña tended to follow each other, forming an ENSO cycle, while the CP El Niño and La Niña tended to occur episodically. Ashok et al (2017) also prove that the two types of El Niño show different evolutionary features of sea surface temperature (SST) and thermocline depth anomalies. The present study examines 21 El Niño events during 1958–2017, showing that the CP El Niño is less likely to evolve into La Niña than the EP El Niño (Figure S1, Table S1).…”

On the Relation between the Boreal Spring Position of the Atlantic Intertropical Convergence Zone and Atlantic Zonal Mode

“…It is well known that the interannual variability of the Indian summer monsoon rainfall (ISMR) is governed by both internal dynamics and external factors such as El Niño-Southern Oscillation (ENSO) (Sikka 1980;Keshavamurty 1982;Mooley and Parthasarathy 1984;Philander et al 1989;Webster et al 1998) and the Indian Ocean dipole (IOD)/zonal mode (IODZM;Behera et al 1999;Ashok et al 2001;Slingo and Annamalai 2000), the dominant interannual modes in the tropical climate. The Atlantic zonal mode (AZM), akin to but weaker than ENSO (Zebiak 1993), is active during boreal summer (June-August) and contemporaneous with the Indian summer monsoon (ISM).…”