In QGIS 2.18.18 on Windows 10, I have downloaded the NASA file srtm_05_01.tif that is available at http://dwtkns.com/srtm and the airports.shp file available at http://qgis.org/downloads/data/qgis_sample_data.zip. I am using them in an exercise to calculate zonal statistics.

I follow these steps to create my analysis area buffer layer and then calculate the zonal statistics

  1. Add the airports.shp layer and the srtm_05_01.tif layer to the project
  2. Vector/Geoprocessing/Tools/Variable Distance Buffer
  3. Accept the defaults and run
  4. Save the resulting buffer layer as a shapefile with the CRS set to EPSG:4326 WGS84, the same as the raster file and project
  5. With the zonal_airport layer created, run Raster/Zonal_Statistics
  6. Select srtm_05_01 as raster layer and zonal_airport as polygon layer, elev as the prefix
  7. Select mean, min, max as the stats
  8. Run.

When I look at the added attributes in zonal_airport, I get the following (only 4 of the 76 Alaska airports are in the srtm_05_01 sector)

enter image description here

Except for Port Heiden, it first appeared that the elevmax in the results can be mutiplied by 3 to get ELEV original attribute (perhaps unit of measure diffence: feet and meters or yards). But with Port Heiden it is apparent that I may have missed something.

What am I doing wrong?

  • 1
    It's not clear what you think the problem is. ELEV is probably feet; SRTM data are in meters, and they don't agree exactly (hence the factor between them for each of the airports is slightly different). I don't see what's wrong with PORT HEIDEN. – Jon May 19 '18 at 20:19
  • If ELEV is in feet and SRTM is in meters, then Port Heiden translates to 87 feet vs the 78 in the shapefile. That is a 10% discrepancy, which seems high. – user51749 May 19 '18 at 22:05
  • 1
    Here's a reference for the vertical accuracy of SRTM:sciencedirect.com/science/article/pii/S2090447917300084 You can see from the abstract that errors of 5m or more are not uncommon. If you want better accuracy, I believe this might help: catalog.data.gov/dataset/… although I don't know the vertical errors. But my point is that there are no errors in your analysis--it is only as good as the underlying data! – Jon May 19 '18 at 22:14
  • Thank you, I'm still getting a sense for the data in geoprocessing and with potential errors of 5m you are definitely correct. – user51749 May 20 '18 at 2:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.