Biases in ENSO response to greenhouse warming induced by ensemble-mean detrending in large ensemble simulations

Abstract Quantifying projected changes in the El Niño–southern oscillation (ENSO) requires an appropriate removal of greenhouse warming-induced trend in associated sea surface temperature (SST) field. Multi-member ensemble average over single-model large ensemble (LE) simulations is frequently used to isolate an externally-forced climate response from internal variability, but whether it can cleanly extract the forced ENSO response to greenhouse warming remains unclear. Using eight LE models, here we show that compared to low-frequency fitting methods that can be applied directly to observation and a single-member simulation, the ensemble-mean detrending approach induces spurious SST variability in the equatorial Pacific, resulting in an underestimated ENSO amplitude response to greenhouse warming. The source of spurious variability relates to a residual effect associated with ENSO nonlinear rectification, which is inherently unavoidable though its influence decreases with ensemble size. Our finding highlights the importance of ENSO nonlinearity in choosing specific ways of detrending for ENSO-related projection studies.