The effect of El Nino on atmospheric processes is best investigated in models where one has the ability to specify sea surface temperatures with and without the El Nino signal. Unfortunately, it has been difficult to eliminate the ENSO signal from historical data sets without eliminating the signal in the entire interannual frequency band, whether or not all of that signal is associated with El Nino. We present a method, based on the statistics of tropical IndoPacific sea surface temperature data, for identifying and removing most of the evolving El Nino signal while retaining the broadband frequency signal not associated with ENSO. We discuss the generation of long synthetic SST time series having the same spectral characteristics of the filtered data and the applications such synthetic data sets might have.