Cross validation uses all of the data to estimate the autocorrelation model. It omits a point and calculates the value of this location using the remaining points in the neighborhood. The predicted and actual values at the location of the omitted point are compared. This procedure is repeated for a second point, and so on until the measured and predicted values of all points are compared.
Note: In a sense, cross validation cheats a little by using all of the data to estimate the autocorrelation model. Hence some prediction may be set aside as unusual, requiring the autocorrelation model to be refit.