# Removing every second row and column cells from a raster file

I have a SRTM 3 arc second topography model (.tif file). If I remove every second row and column of cells from the grid, can it be said that the model becomes a 6 arc second?

Or does SRTM 3 arc second model needs some kind of interpolation of its grid in order to get the 6 arc second model? If it does need, can anyone provide GDAL string example of how this can be done?

• No, you have to resample it to new cell size Commented Mar 8, 2016 at 22:12
• @FelixIP Removing every other row and column is resampling (it's an instance of the nearest-neighbor algorithm). Commented Mar 8, 2016 at 23:05
• Thank you for the replies both FelixIP and whuber. @whuber: So can a 3 arc second model be considered to be a 6 arc second model if every other row and column is removed? Commented Mar 8, 2016 at 23:17
• Of course--the spacing is now twice 3 arc-seconds, right? For many purposes this approach (nearest-neighbor resampling) isn't a great choice, though. You might consider availing yourself of other resampling options. Cubic convolution is attractive in this setting. Commented Mar 11, 2016 at 16:08
• Thank you @whuber. I took a look at both the linked topic and your article at quantdec.com. Yes, the spacing is twice 3 arc-seconds, and it has about 4 times less cells than 3 arc-seconds. So "manual" (sort to say) interpolation in this way (removing every second row and column of cells from the grid) can basically produce a 6 arc-seconds DEM from 3 arc-seconds DEM? This is what I want to know. Regardless if this approach is more appropriate/precise than using some other resampling methods (like cubic convolution). Commented Mar 11, 2016 at 17:45

``````gdal_translate -outsize 50% 0 input.tif output.tif
Read the manual http://www.gdal.org/gdal_translate.html because otherwise you can't know what `-outsize 50% 0` does nor that `-r` is not needed because nearest is the default resampling algorithm.