How Sample works

If the input locations are specified as a point feature dataset, the point coordinates may not be the same as the cell centers of the cells specified in the input raster. To determine the values at the exact locations of the selected points, a resampling technique must be adopted. For discrete data types, the default NEAREST resampling algorithm will be satisfactory. However, for continuous data types, it is more appropriate to perform another resampling technique. For instance, for elevation data, applying bilinear interpolation or cubic convolution will provide more accurate results. The bilinear option will determine the cell values based on a weighted distance of the four nearest input cell centers. The cubic option will compute the cell value based on the weighted distance of the nearest 16 cells.

For each sample location, information on the location number (the raster value or point feature ID), location in map space (x,y coordinates), and cell values from each of the input rasters is written to the table. The table has the following general structure:

   locationID-1 x-coord1 y-coord1 cellvalue1 cellvalue2 cellvalue3 ....
   locationID-2 x-coord2 y-coord2 cellvalue1 cellvalue2 cellvalue3 ....

An example of an FGDB table output from the Sample tool follows: