I have a floating point raster having dimensions of 478 x 418 columns and rows respectively. The cell size is 30 x 30 meters, therefore, it covers an area of 179.5 Sqkm.

the landslide probability Raster

The intent is to obtain one row for each pixel of the map in a table.

Is there any way around except 'Raster to Points' tool available in ArcMap?

As such, I would like to retain values as pixels to compute other statistics required for validation of this raster through Success Rate and Prediction Rate Curves! Like, frequency of training/testing points in each cell.

How should I reclassify it by retaining its original values in contrast to assigning new values using 'Reclassify tool?


2 Answers 2


you probably don't have training/testing points for each pixel. Therefore you should extract the values of the pixels for each testing point (using extract value to point or extract multivalue to point) instead of converting your raster pixels to unique value.

  • to calculate success rate and prediction rate curves, I have set of points but surely do not overlay each pixel. The values of the floating point raster are stretched from low to high based on training datasets used in the analysis.
    – Dave
    Commented Jan 3, 2017 at 9:28

You can build a raster table directly using http://desktop.arcgis.com/en/arcmap/10.3/tools/data-management-toolbox/build-raster-attribute-table.htm

However this won't work on 32bit floating point data. If you need the decimal places, multiply by e.g 1000 then convert to integer first.

  • Build raster attribute tool works only for integer raster datasets. Transforming my values from floating to integer is not desired because it truncates my values stored as decimal places.
    – Dave
    Commented Jan 3, 2017 at 9:26
  • Secondly, even if I divide the resultant values (after multiplying with 1000) it removes all of my values having -signs, which is vital.
    – Dave
    Commented Jan 4, 2017 at 4:09

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.