Uncertainties in SST datasets and implications for El Niño event classification

Identification of El Niño warm events and their types has traditionally been deterministic, based mainly on whether a pre-defined index exceeded a critical value. However, uncertainties in both sea surface temperature (SST) measurements and their interpolation into a gridded analysis can impact identification of and confidence in El Niño variability, particularly earlier in the record. Although several different classification methods for El Niño exist, researchers lack an effective reference and evaluation system to identify advantages and disadvantages of a given index for a given application. Therefore, this study quantifies the impacts of both data- and method-related uncertainties on different El Niño classification methods, considering different types of uncertainty, different types of analysis, different teleconnection mechanisms and expressions of El Niño impact and different types of climate data. To aid in these objectives, El Niño classification methods are evaluated from five aspects: reliability, accuracy, precision, flexibility, and simplicity. The core analysis is based on probabilistic, uncertainty-aware classifications applied to a large ensemble of historical SST realizations. The results are then used to conduct a more general evaluation of how different types of uncertainty propagate through the different classification methods, and provide guidance on the strengths and weaknesses of these indices for different applications.