Clustering 30m pixels into 120m pixels, fishnet using QGIS?

Question

I have a raster file representing tree cover loss made of 30m*30m pixels taking the value 0 or 1. 0 means no loss and 1 means tree cover loss. I want to cluster these pixels into bigger pixels of the size 120m*120m (i.e. 4 pixels by 4 pixels so 16 pixels in each cluster). I plan to do this for computational reasons as by doing so I reduce the number of observations by a factor of 16. My goal is to measure the share of 30m pixels with value 1 within each 120m cluster of pixels.

I am not sure how to do this.

Here is a solution I thought of:

• On QGIS I thought of creating a fishnet of 4 by 4 pixels and then doing zonal statistics within each cell. BUT, I could not find how to create this 4 by 4 pixels grid AND I found no solution yet on how to match perfectly cell boundaries with pixel extent (in order to have exactly 16 pixels per fishnet cell, see red cells below).

Answer 1

So I found a second best solution to the question asked above.

Using R one can aggregate raster pixels by a factor, in my case a factor of 4, and then export the newly created raster. This raster can subsequently be used in QGIS.

Please note that this can take a very long time. My raster is a 80000*40000 pixels so aggregation takes about one hour per raster file.

library(raster)
DEM <- raster("/Users/.../input.tif")
dem_low <- aggregate(DEM, fact = 4, fun = mean)
writeRaster(dem_low,'/Users/output.tif',options=c('TFW=YES'))
rm(DEM,dem_low)

Answer 2

GDAL is a great tool for re-sampling rasters as well, for example using gdal_translate:

gdal_translate -tr 120 120 -r average input.tif output.tif

https://www.gdal.org/gdal_translate.html