I am assessing the accuracy for the 2011 National Land Cover Map (NLCD). Each pixel for the NLCD measures 30m x 30m. For my reference data, I'm using two sources:

1) Google Earth (~20 m resolution) 2) NAIP hi res imagery (1-2 m resolution)

For each randomly sampled point in my hi res imagery, I want to create a grid that lays over it and simulates the area of the pixel it represents on the NLCD (30m x 30 m). This will allow me to label the land cover based on the rules for the NLCD classification scheme (if more than 30% of the 30 m x 30 m pixel is covered in leaf foliage, it is forest).

Does anyone know of a tool for creating a grid like this in Arc that works on raster? The create fishnet tool is for vector.

  • One would suppose that a good accuracy assessment would run the other way: for each (randomly selected) cell of the NLCD, collect all the values of intersection cells from the reference data (GE or NAIP). Compare a statistical summary of those matching cells to the value in the NLCD cell. – whuber Mar 25 '16 at 21:12
  • Yes, I created the random points for each land cover of the NLCD (the sample size for each was based on the land cover's area relative to the total area) . Then I converted the points to kml and pulled it up in GE and NAIP – Ashley Lynn Tarver Mar 26 '16 at 22:33

There are several ways to do that. You could create a fishnet with proper spatial information first. Then convert it to a raster based on the FID field of the fishnet. The reason for using the FID field is that the FID value is unique for each cell, so when you convert the shapefile there is no chance that adjacent cells would merge, resulting a standard grid that you want.

  • I don't understand the use of the FID. The cell is identified by its spatial location. I'd think you would use a value that represents your determination for that cell. – jbchurchill Mar 25 '16 at 12:57
  • @jbchurchill The conversion from polygons to raster in ArcGIS is based on the values of a certain field. Of course you could use spatial information, but that is an extra step you need to take. However, the FID field is there readily. – TonyC Mar 25 '16 at 16:34
  • Actually, now that I'm thinking about it, I don't think I need to convert it from vector to raster at all, that would compromise the precision, and precise measurement is the reason for needing the fishnet in the first place. I can just lay the fishnet (measuring 30m x 30 m) over the GE imagery (measuring 15 m) and account for the difference in resolution between the two – Ashley Lynn Tarver Mar 26 '16 at 22:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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