Are the previous ENSOs being used to train the models? Or are the projections just results of model ensembles given current inputs without reference to the previously recorded anomalies?