0

I was given one single band NDVI raster and a shapefile with multiple very small polygons.

The raster has a resolution of 3m. On the other hand, the small polygons on the shapefile are square shaped with each side 2m long. Thus, any pixel in the raster is larger than the polygon on the shapefile.

I was told to extract the raster value of each polygon. After running the Zonal Statistics as Table I experienced the following issues -

  1. When the small polygon completely falls into one single pixel as shown in the image below. Here the pixel has a value of 0.5. After zonal statistics, the value of the polygon is also shown as 0.5. This is completely fine.enter image description here

  2. Sometimes the small polygon falls into multiple pixels as shown in the image below. Suppose, the upper half of the polygon falls into a pixel with 0.3 value and the lower half falls into a pixel with 0.4 value. I was expecting that after the zonal statistics as table operation, I get a value of 0.35 (.5*.3+.5*.4). However, it does not provide any values for this polygon. The polygon that falls into multiple pixels does not even appears in the resulted table.

Is there any way of solving the second point? I am currently using ArcMap 10.5.

enter image description here

Sourav
  • 257
  • 1
  • 13
  • Related, but for QGIS (although I'd expect some similarities): https://gis.stackexchange.com/questions/276794/how-does-qgis-zonal-statistics-handle-partially-overlapping-pixels – Jon Jul 31 '18 at 20:57
  • 1
    Find centroids and sample raster. – FelixIP Jul 31 '18 at 21:05
  • @FelixIP can you please explain a bit? – Sourav Jul 31 '18 at 21:54
  • Add geometry attributes, centroid. Make xy layer. Export to points. Extract values, spatial analyst. Join back to polygons if necessary. – FelixIP Jul 31 '18 at 22:12
  • @FelixIP has the right idea, you could alternately use Feature to Point http://resources.arcgis.com/en/help/main/10.2/index.html#/Feature_To_Point/00170000003m000000/ if you have an advanced license, but I'd do this extra step only for the polygons that have values missing.. you still want to get Zonal Statistics of the larger polygons covering more than one cell. The reason for this aberration is because the Zonal Statistics tool internally rasterizes the polygons by FID, if your polygon fails to fill a cell then there is no valid pixel in the intermediate raster = no output row. – Michael Stimson Jul 31 '18 at 23:06
  • @FelixIP I had the point shapefile and extracting just the single value of a point is not helping me for my analysis. I have to extract the weighted average (based on the portion of the polygon they are talking) of the NDVI value of all pixel of the raster falls inside the polygon. – Sourav Aug 01 '18 at 15:47
  • 1
    It means that your weighted average equal value at this point. Go and tell them they don't understand how raster work or aternative: connvert raster to polygons and intersect. You'll get slightly different result = waste of time. – FelixIP Aug 01 '18 at 18:41
  • You can resample the raster and make it a finer resolution. – M-- Oct 27 '23 at 14:52

0 Answers0