I have a point shapefile that includes parcel centroid features across a county in Pennsylvania.

I am trying to get a sense of the percentage of forest land within half-mile and mile buffers around each of the centroids. Using NLCD land cover data, I've created a raster that indicates the presence of forest (== 1).

I've then applied Focal Statistics to that raster to create a raster that indicates the % of area within a 1 mile buffer of each cell that is in forest.

I've then tried to apply these raster values to the parcel centroid shapefile using Zonal Statistics by table.

Unfortunately, it hasn't worked. Out of 28000 centroids, it's only working for 25000 of the points.

Why might this be happening?

  • 1
    How exactly are you "applying" the raster values to a point shapefile? If you're trying to use the points as individual zones, first convert them to raster format and check that you get as many unique values in the raster's VAT as there are points. (You likely will not whenever centroids occupy a common raster cell.)
    – whuber
    Mar 23, 2015 at 20:00
  • If you area talking parcels (not their centrioids) it very well might be a small (relative to cell size) parcels that are missing.
    – FelixIP
    Mar 23, 2015 at 21:11
  • What did you determined to be the zone value? zonal stats apply statistics to zones by their value; that is if you had duplicate values in a field, and if that field was use as the zones field - than they were aggregated together; you can check it by summrise the area in the zonal output table and compare with the total area of the relevant buffers. Alternatively, you can apply zonal stats again, and make sure you use Unique field, e.g. FID, as the zone field.
    – dof1985
    Mar 23, 2015 at 21:29
  • 1
    Are you sure you're not trying to extract values by point? If you've created a raster that for cell values has percentage of forest cover within 1 mile of that cell, and then trying to get that as an attribute of the points, you're using the wrong tool. Zonal Stats would be what you would run on say, buffer polygons sitting on the original landcover raster.
    – Chris W
    Mar 23, 2015 at 21:35
  • 1
    Is the 'not forest' NoData? You wont get statistics for points that fall entirely in NoData. You should be performing your statistics using a buffer of the centroids (can overlap). Mar 24, 2015 at 0:08

1 Answer 1


Note the following from the 'Zonal Statistics as Table' tool; "If the zone input is a point feature dataset, it is possible to have more than one point contained within any particular cell of the value input raster. For such cells, the zone value is determined by the point with the highest feature ID". I suspect that the points that 'failed' may have been those of which other points with higher feature IDs occur within the same cell?

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.