I have a few NumPy arrays: lat, lon, and 6 data bands. I'd like to convert these to GeoTiff and then apply the gdalwarp function, based on a specific projection, to each band.

Is there a standard way to do this in Python?

  • possible dublicate: gis.stackexchange.com/questions/37238/…
    – Thomas
    Jun 11, 2016 at 22:27
  • Possible, yes. A major difference is that my lat/lon resolutions aren't always uniform. Defining the geotransform is the key, I think.
    – rakbar
    Jun 11, 2016 at 23:04
  • So they don't form a complete grid? In this case you will either have to interpolate a surface from the observations, or be left with missing values. Any preference? Can you better characterize the presence of lat/lon observations for us?
    – Jeff G
    Jun 12, 2016 at 14:34
  • 1
    To give you more context: I have a series of HDF5 files, with swath grided data and lat/lon values. The coordinate description is " geodetic lat/lon of the center of each 1 km cell." I need to first apply some flagging and filtering operations to the data, then re-project to EPSG:6933. I was making virtual rasters (.vrt files) then using gdalwarp to do all this. But because of the flagging operations, things got a bit complicated, hence numpy arrays. The HDF5 don't have an specific geolocation meta data either. Hope this helps!
    – rakbar
    Jun 12, 2016 at 22:11

1 Answer 1


Can be done fairly easy. First, create the dataset:

drv = gdal.GetDriverByName("GTiff")
ds = drv.Create("name.tif", width, height, 6, gdal.GDT_Float32)

You then need to set the geotransform using ds.SetGeoTransform where the argument is a six element tuple: (upper_left_x, x_resolution, x_skew, upper_left_y, y_skew, y_resolution)

Set the projection using ds.SetProjection, argument is the way string, then Write the array for the first band:


That should be enough to get you going. If the purpose is just to reproject the data, then you could look at using an in memory dataset

  • 2
    Make sure to use ds.SetGeoTransform instead of ds.SetGeotransform
    – ArnJac
    Feb 14, 2018 at 15:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.