I have been attempting to use the Spatial Analyst Curvature tool with mixed results. I originally tried to use it with a raster that was projected using a 1983 Teale Albers projection and the results were black, indicating the lowest values. I then transformed the raster into UTM projection and ended up with grid-like lines on the raster which have values that are obviously not accurate. Does anyone know why these lines have appeared or what I'm doing wrong. Thanks.

enter image description here

  • Those might be compression artifacts. Were the data at any point stored in JPEG format, perchance? Is this image a semi-transparent overlay of the DEM and a curvature layer or is it only displaying the top (curvature) layer? – whuber Dec 6 '13 at 18:33
  • Thanks Gentlemen. I tried reprojecting it into multiple different projections using the bilinear resampling but have only resulted in negative infinity values (black) when using the tool now. Any suggestions? I've been trying different UTM projections including NAD83 10N, NAD83 HARN 10N, and NAD83 (2011) 10N. My site is near Etna, CA and I'm fairly certain that I'm in the right area and the linear distance is in meters and not decimal degrees like as I found as a problem on a different page. Thanks again!!! – sdb348 Dec 7 '13 at 1:51
  • After projecting it try calculate statistics on the raster. ArcGIS should do this automatically but does not always. In ArcCatalog right click on the raster and select Calculate Statistics... This may fix the black raster issue. – Jeffrey Evans Dec 7 '13 at 15:29

I believe that these artifacts are because when you reprojected the DEM you defaulted to a nearest neighbor resampling. Try reprojection the data using bilinear resampling. SOme of the additional artifacts are likely due to you using a 10m DEM from NED. This data has terrible contour biasing and leads to linear artifacts in derivatives, like you are seeing in the drainage patterns.

| improve this answer | |
  • 1
    +1 I think you're correct about the nearest-neighbor resampling: that will create regular lines of doubled cells, to which curvature is particularly sensitive. The contoured look of the DEM (if indeed that's what we're seeing) also supports this conjecture: all elevations must be integral multiples of a common unit. – whuber Dec 6 '13 at 18:45
  • Thank you Mr. Evans. I found the original .tif files and then calculated the statistics. This worked like a charm. The file was in the NAD87 10N projection. After I got it to work, I realized that I never reprojected it with bilinear resampling. Regardless, those artifacts were gone. Thanks again. – sdb348 Dec 8 '13 at 0:43

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.