Evaluation of Time Series Causal Detection Methods on the Influence of Pacific and Atlantic Ocean over Northeastern Brazil Precipitation

The detection of causation in natural systems or phenomena has been a fundamental task of science for a long time. In recent decades, data-driven approaches have emerged to perform this task automatically. Some of them are specialized in time series. However, there is no clarity in literature what methods perform better in what scenarios. Thus this paper presents an evaluation of causality detection methods for time series using a well-known and extensively studied case study: the influence of El Niño-Southern Oscillation and Intertropical Convergence Zone on precipitation in Northeastern Brazil. We employed multiple approaches and two datasets to evaluate the methods, and found that the SELVAR and SLARAC methods delivered the best performance.