A Machine Learning-Based Approach to Quantify ENSO Sources of Predictability

A machine learning method is used to identify sources of long-term ENSO predictability in the ocean (sea surface temperature (SST) and heat content) and the atmosphere (near-surface zonal wind (U10)). Tropical SST represents the primary source of predictability skill. While U10 does not increase the skill when associated with SST, our analysis suggests U10 alone has apredictive skill comparable to that of SST between 11 and 21 months in advance, from late fall up to late spring. The long-lead signal originates from coupled wind-SST interactions across the Indian Ocean (IO) and propagates across the Pacific via an atmospheric bridge mechanism. A linear correlation analysis supports this mechanism, suggesting a precursor link between anomalies in SST in the western and wind in the eastern IO. Our results have important implications for ENSO predictions beyond 1 year ahead and identify the key role of U10 over the IO. Many extreme events, such as floods or droughts, can be attributed to the El Ni & ntilde;o Southern Oscillation, a mode of large-scale ocean-atmosphere coupled variability in the tropical Pacific Ocean occurring with a period of approximately 4 years. In this analysis, we use a machine learning methodology to disentangle the key atmospheric and oceanic ENSO components' relative contribution to its predictability, particularly the role of near-surface 10-m zonal wind. We quantify the potential for improved ENSO predictions for up to 2 years in advance and present a mechanistic understanding of the location of the sources of predictability. While equatorial sea surface temperature represents the primary source of ENSO predictability, the equatorial U10 plays a vital role from late spring to fall, from 1 to 2 years in advance. The enhanced predictability skill is shown to be linked to an SST anomaly originating in the Indian Ocean. The ML model used provides a new way to get new insights into the sources of predictability for ENSO and can be used as a simple but powerful tool to improve the underlying mechanistic understanding. A deep learning-based approach suggests near-surface 10 m wind to play a significant role in providing skills for long-term ENSO predictability The skill of the 10-m zonal wind is large from 12 months up to 23 months in advance (depending on the season) The signal is generated by coupled wind-SST interactions in the Indian Ocean and later propagates across the Pacific