Local Trend

The trend you chose to remove drift in Trend Surfaces, was applied to the entire study area. To begin cross validation you must define what trend exists on the scale of estimation, that is, the local trend. Choose a trend from the Local trend list.

If a high order of trend e.g. cubic, is fitted to the study area, it may well be possible, and indeed necessary, to reduce the order by one or two when working at the estimation scale. That is, reduce a cubic trend to quadratic or linear. You can do this because the estimated value for any particular point is based on data points in the immediate vicinity, not all the points in the data set.