3

When using the reduceNeighborhood function with a kernel, the results differ according to the zoom level. When the results are zoomed in at the maximum (you see individual pixels clearly), the kernel is computed only for the edges of a pixel. The differences are also evident when interrogating the same pixels over multiple zoom levels. See below the results from the GEE example from Statistics of Image Neighborhood.

Are there some parameters to setup in order to have the same results no matter what the zoom level is?

Zoom out image

zoom_out

Zoom in image

zoom_in

2

You can force an analysis to be performed at a certain scale using the ee.Image.reproject(crs, crsTransform, scale) method.

There are implications of using forcing your analysis to use a particular scale when analyzing large geographic areas. See the Reprojecting section of the Earth Engine docs for a full discussion.

  • Just to add to this, it's worth noting that the "edges" you're seeing are in fact the result of the input being resampled into pixels smaller than the native scale (where output scale is set from the zoom level). All those little pixels are input to the reducer, so in the middle of the big pixels there are homogenous values while along the edges there is positive variance. – Nicholas Clinton Jan 5 '18 at 1:59

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.