4

This question is hard to explain so if you have a possible answer I would love to hear it.

I work in the precision agriculture business which heavily relies on NDVI calculations but need to create like zones of crop health for field surveyors and field management purposes. The end users for maps I create are not GIS savvy so presentation is everything. Raster Classification, zonal fills, and reclassification's are not what I want because they are vague and distort the data. On the attached map there are geo-coordinate points over areas with possible health strains and areas that appear to be very healthy.

How do I create zone lines, almost like contour lines, to batch similar points into like areas?
For example, in the Northwest portion of the map, the three points are over similar areas. How do I create a boundary lines showing different zones?
I'm looking for 5 areas:
i) worst;
ii) bad;
iii) OK;
iv) good;
v) best,
separated by contour like lines creating zones. I don't want to disrupt are high resolution photography.

enter image description here

  • 1
    Welcome to GIS.SE! You don't need to worry about getting trolled here. What version of ArcMap are you using? – Paul Aug 8 '13 at 18:59
  • Arc 10.1 but upgrading to 10.2, its better equipped for NDVI analysis supposedly. We'll see – FadeLincoln Aug 8 '13 at 19:41
  • One issue is that there appears to be "relics" in your image, the wheel tracks of machines appear a bit darker. With 5 zones will these be captured in the wrong class? A useful part of your workflow would be to create a histogram and then see where the important values sit on the map. – Willy Aug 13 '13 at 22:02
2

I might approach this in two steps:

First, I would use the raster reclass tool to classify the values as high, medium, or low. Use your expertise to judge the break points on these.

Then, depending on how much you want to reduce noise, I would generalize the data using a majority filter. That will create more homogenous zones. It might mess with your dendritic areas, though, so proceed with caution.

The next step is to use boundary clean to clean up the edges of your data. You can further generalize your classifications by using region group, set null, and nibble. For more on this workflow, see http://resources.arcgis.com/en/help/main/10.1/index.html#/Processing_classified_output/00nv00000015000000/

  • 1
    The problem I have with filters and zones is that it will generalize the perceived health of the crop. I'd like to create zones then come up with percentages of each health level in the areas and generalize that way. ie: Zone "X" contains 50% ok crop, 30% bad, and 20% excellent – FadeLincoln Aug 8 '13 at 19:46
  • In that case, you'll have to define your zones manually, then take the reclassified raster and split it into the zones. It would then be pretty easy to calculate percentages for each zone manually. Again, I defer to your specialized expertise as to what makes sense for each zone. Some ideas include field boundaries, soil survey polygons, or buffers around the channels and field edges. – Jay Guarneri Aug 8 '13 at 23:12
  • 1
    It seems as though you're somewhat against geo-statistical processing because you don't wish to lose the realism/texture of the original image? Can you not "simply" do the analysis and overlay it's results transparently over the original, producing the best of both map effects? – Martin F Aug 15 '13 at 20:35

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.