Glossary for Geospatial Science

nearest neighbor sampling/interpolation:� Resampling a source raster to yield a new raster with a different cell size, raster orientation, and/or internal geometry by computing the distance between the center of each cell in the output raster and the 4 nearest cells in the input raster.� The data value for the closest input cell is assigned without alteration to be the data value of the output cell.� Therefore, the input value of one input cell may be assigned to more than one output cell.� It also means that some input cells may not be transferred at all to the output raster.� These undersampling and oversampling situations occur when the cell sizes of the input and output raster are different.

For example, a resampled output raster which is coparallel to the input raster but which has cells half as big, will repeat the input values in a 2 x 2 cell pattern.� However, since nearest neighbor resampling does not mathematically compute the new cell�s value, it is the only suitable choice for rasters that are not mathematically continuous such as land cover maps.� With such maps, mathematical resampling (like bilinear, or cubic convolution) is totally incorrect since mean values cannot be computed.