1

I know this is not a new question, and I know it depends on the context and what I would like to do with the data, but I am still finding it difficult to solve this problem in practice. And I haven't found any good answer online (this question was marked as possible duplicate, but I do not think this is the case, so I have modified my question, I hope it is clearer now).

For example: I have a raster image of modeled atmospheric N deposition for the whole world, with a resolution of 5 by 3.75 degrees. Now I would like to resample it to a finer resolution, say with cubic resampling from GDAL warp.

Atmospheric N deposition, World

EDIT: The reason I want to do this is that I would like to estimate N deposition per country. I understand I am introducing an error when I do any resampling, however as you can see in the image, the pixels are really quite large compared to the size of an average country.

Now I could run a zonal statistics on a world countries vector layer, using this raster image as it is. However, I suspect that my results would be of better quality if I resample the raster image first, to a higher resolution. If I am wrong, than please let me know and I will close this question myself.

Question: Can you please explain what logical steps and reasoning I should follow to decide the new resolution of the raster image? The aim here is to then run a Zonal Statistics on the higher resolution raster image, with a vector layer containing country boundaries.

I would like this to be an example of how to make such a decision, I don't care so much about the answer to this specific case.

ffgg
  • 113
  • 7
  • 1
    Please focus your post by including a single question. – Aaron Feb 12 '20 at 13:48
  • Thank you for your comment, I've edited the question, and split the question into two parts, is this better? – ffgg Feb 12 '20 at 13:56
  • 1
    @ffgg: There's a one question per post-policy, please try to comply with it. Also, your question is rather well researchable, maybe try that first and then ask more specific questions. – Erik Feb 12 '20 at 14:04
  • 1
    Thanks for the edits, however, I still count 5 questions. This post is likely to be closed by the community unless you can focus in one question. – Aaron Feb 12 '20 at 14:05
  • 1
    Thank you, I have now focused in one question, I hope this is good enough – ffgg Feb 12 '20 at 15:00
  • 2
    As a general rule you cannot resample a raster to create new data. For example, you cannot resample a precipitation raster to produce more accurate localized data. Resampling to a finer resolution is usually done to align different datasets for further analysis. – Aaron Feb 12 '20 at 15:13
  • 1
    @Aaron, answered your question: the new “artificial” resolution will be based on other datasets (if they exist) if they are needed. Avoid using bilinear interpolation for resampling (prefer nearest neighbor in this case); see https://gis.stackexchange.com/questions/27838/resample-binary-raster-to-give-proportion-within-new-cell-window/27849#27849. – Andre Silva Feb 12 '20 at 15:22
  • I see you have marked my question as a duplicate but I don't think the answer in the linked question is very relevant. Can I edit my question further or should I open a new question? – ffgg Feb 12 '20 at 16:24
  • 1
    You should edit your question and explain at the beginning why you think that it is not a duplicate. – PolyGeo Feb 12 '20 at 19:52
  • 2
    I agree with @PolyGeo, please edit your question to indicate how yours is different from the linked duplicate. I see the answer in the duplicate as a good solution to your question. – Aaron Feb 12 '20 at 21:30
  • Ok thank you all for your feedback and patience, I have edited my question further. – ffgg Feb 13 '20 at 10:37

0 Answers0