Classifying El Niño-Southern Oscillation Combining Network Science and Machine Learning

Machine learning and complex network theory have emerged as crucial tools to extract meaningful information from big data, especially those related to complex systems. In this work, we aim to combine them to analyze El Niño Southern Oscillation (ENSO) phases. This non-linear phenomenon consists of anomalous (de)increase of temperature at the tropical Pacific Ocean, which has irregular occurrence and causes climatic variability worldwide. We construct temporal Climate Networks from the Surface Air Temperature time-series and calculate network metrics to characterize the warm and cold ENSO episodes. The metrics are used as topological features for classification. We employ ten classifiers and achieved 80% AUC ROC when predicting the intensity of Strong/ Weak El Niño and Strong/Weak La Niña for the next season. The complex network represents the relationship among different regions of the planet and machine learning creates models to classify the different classes of ENSO. This work opens new paths of research by integrating network science and machine learning to analyze complex data like global climate systems.

[1]  Y.-Y. Liu,et al.  The fundamental advantages of temporal networks , 2016, Science.

[2]  Jürgen Kurths,et al.  Complex networks reveal global pattern of extreme-rainfall teleconnections , 2019, Nature.

[3]  R. Ren,et al.  Parallel comparison of the 1982/83, 1997/98 and 2015/16 super El Niños and their effects on the extratropical stratosphere , 2017, Advances in Atmospheric Sciences.

[4]  S. Koopman,et al.  Long-term forecasting of El Niño events via dynamic factor simulations , 2020, Journal of Econometrics.

[5]  J. C. Furtado,et al.  The role of South Pacific atmospheric variability in the development of different types of ENSO , 2017 .

[6]  Jari Saramäki,et al.  Temporal Networks , 2011, Encyclopedia of Social Network Analysis and Mining.

[7]  Michael Small,et al.  Brain anomaly networks uncover heterogeneous functional reorganization patterns after stroke , 2018, NeuroImage: Clinical.

[8]  B. Kirtman,et al.  The South Pacific Meridional Mode as a Thermally Driven Source of ENSO Amplitude Modulation and Uncertainty , 2018, Journal of Climate.

[9]  Liang Zhao,et al.  From spatio-temporal data to chronological networks: an application to wildfire analysis , 2018, SAC.

[10]  S. Havlin,et al.  Climate networks around the globe are significantly affected by El Niño. , 2008, Physical review letters.

[11]  S. Xie,et al.  The North Pacific Pacemaker Effect on Historical ENSO and Its Mechanisms , 2019, Journal of Climate.

[12]  Emily A. Fogarty,et al.  Visibility network of United States hurricanes , 2009 .

[13]  Ryan L. Sriver,et al.  Analyzing El Niño–Southern Oscillation Predictability Using Long‐Short‐Term‐Memory Models , 2019, Earth and Space Science.

[14]  Timothy J. Hoar,et al.  The 1990–1995 El Niño‐Southern Oscillation Event: Longest on Record , 1996 .

[15]  Alneu de Andrade Lopes,et al.  A Comparison of Graph Construction Methods for Semi-Supervised Learning , 2018, 2018 International Joint Conference on Neural Networks (IJCNN).

[16]  C. Garfinkel,et al.  Modulation of the Northern Winter Stratospheric El Niño–Southern Oscillation Teleconnection by the PDO , 2019, Journal of Climate.

[17]  Jeong-Hwan Kim,et al.  Deep learning for multi-year ENSO forecasts , 2019, Nature.

[18]  Evangelos E. Milios,et al.  A multi-centrality index for graph-based keyword extraction , 2019, Inf. Process. Manag..

[19]  R. Reynolds,et al.  The NCEP/NCAR 40-Year Reanalysis Project , 1996, Renewable Energy.

[20]  G. Stenchikov,et al.  Impacts of a Pinatubo‐size volcanic eruption on ENSO , 2017 .

[21]  Mark Newman,et al.  Networks: An Introduction , 2010 .

[22]  Kamiar Mohaddes,et al.  Fair Weather or Foul? The Macroeconomic Effects of El Niño , 2015 .

[23]  Shlomo Havlin,et al.  Improved El Niño forecasting by cooperativity detection , 2013, Proceedings of the National Academy of Sciences.

[24]  Delphine Clara Zemp,et al.  Node-weighted measures for complex networks with directed and weighted edges for studying continental moisture recycling , 2014 .

[25]  Alneu de Andrade Lopes,et al.  Network Sampling Based on Centrality Measures for Relational Classification , 2015, SIMBig.

[26]  Alneu de Andrade Lopes,et al.  RGCLI: Robust Graph that Considers Labeled Instances for Semi-Supervised Learning , 2017, Neurocomputing.

[27]  Norihiko Shinomiya,et al.  El Niño-Southern Oscillation forecasting using complex networks analysis of LSTM neural networks , 2019, Artificial Life and Robotics.

[28]  Marc Wiedermann,et al.  A climate network‐based index to discriminate different types of El Niño and La Niña , 2016, 1604.04432.

[29]  A. Barnston,et al.  Clustering of eastern North Pacific tropical cyclone tracks: ENSO and MJO effects , 2008 .

[30]  Klaus Lehnertz,et al.  Unraveling Spurious Properties of Interaction Networks with Tailored Random Networks , 2011, PloS one.

[31]  M. Latif,et al.  Greenhouse Warming, Decadal Variability, or El Niño? An Attempt to Understand the Anomalous 1990s , 1997 .

[32]  J. Kurths,et al.  Complex network approaches to nonlinear time series analysis , 2019, Physics Reports.

[33]  S. Havlin,et al.  Network analysis reveals strongly localized impacts of El Niño , 2017, Proceedings of the National Academy of Sciences.

[34]  Marc Wiedermann,et al.  Complex Network Techniques for Climatological Data Analysis , 2017 .

[35]  J. Kurths,et al.  Complexity-based approach for El Niño magnitude forecasting before the spring predictability barrier , 2019, Proceedings of the National Academy of Sciences.

[36]  Lilian Berton,et al.  Evaluating link prediction by diffusion processes in dynamic networks , 2019, Scientific Reports.

[37]  Huan Liu,et al.  Feature Selection for Classification , 1997, Intell. Data Anal..

[38]  Michael K. Tippett,et al.  Deterministic skill of ENSO predictions from the North American Multimodel Ensemble , 2017, Climate Dynamics.

[39]  Ernestina Menasalvas Ruiz,et al.  Combining complex networks and data mining: why and how , 2016, bioRxiv.

[40]  Elbert E. N. Macau,et al.  Global fire season severity analysis and forecasting , 2019, Comput. Geosci..

[41]  M. Gan,et al.  Future increase in extreme El Nino events under greenhouse warming increases Zika virus incidence in South America , 2019, npj Climate and Atmospheric Science.

[42]  A. Barnston,et al.  Skill of Real-Time Seasonal ENSO Model Predictions During 2002–11: Is Our Capability Increasing? , 2012 .

[43]  Ilias Fountalis,et al.  Dimensionality Reduction and Network Inference for Climate Data Using δ‐MAPS: Application to the CESM Large Ensemble Sea Surface Temperature , 2019, Journal of Advances in Modeling Earth Systems.

[44]  R. Allan,et al.  ‘Persistent’ ENSO sequences: how unusual was the 1990-1995 El Niño? , 1999 .

[45]  Mark A. Cane,et al.  Civil conflicts are associated with the global climate , 2011, Nature.

[46]  Thomas M. Smith,et al.  Extended Reconstructed Sea Surface Temperature, Version 5 (ERSSTv5): Upgrades, Validations, and Intercomparisons , 2017 .

[47]  Shlomo Havlin,et al.  Forecasting the magnitude and onset of El Niño based on climate network , 2017, 1703.09138.

[48]  Potsdam,et al.  Complex networks in climate dynamics. Comparing linear and nonlinear network construction methods , 2009, 0907.4359.