It seems likely those large scale variations correlate with summer temperatures, but from the information you give there, that’s n=1, self-evaluated, after removal of values that do not fit the prediction.

Are you sure this can be used to make predictions for single seasons?