On the relative role of east and west pacific sea surface temperature (SST) gradients in the prediction skill of Central Pacific NINO3.4 SST

Abstract

Skills in NINO3.4 sea surface temperature (SST) prediction provide a benchmark for evaluation of the current generation of machine learning models. Several empirical data-driven models rely on capturing low-frequency variability of the SST anomalies over the east and west Pacific as a dominant predictor. The physical processes contributing to the SST anomalies in the east and west Pacific are different. The study discusses the relative contribution of SST anomalies over the western and eastern Pacific to the prediction skill of NINO3.4 SST using a convolutional neural network (CNN)–based prediction model. CNN models employ spatial filters and are highly efficient in capturing the anomaly edges or gradients. The study reports three CNN-based model experiments. The first is a CTRL experiment using the whole equatorial Pacific domain SST as input. The second and third models use the equatorial eastern and western Pacific domain SST only. A novel feature of this study is that we have generated a large number of ensemble members (5000) through random initialization of CNN filters. It is found that random initialization affects the forecast skill, and the skill of model ensembles at each lead time shows a Gaussian distribution. The analysis suggests that the west Pacific SST model provides better NINO3.4 skills as compared to the east Pacific models. The model forecast skills also show monthly variability with low skill during April–May, indicative of the spring predictability barrier. Ensembles with good skill show relatively better east-west gradients than ensembles with bad skill.