Best use of r.out.gdal for large raster

GRASS GRASS user
,
1

Ciao,
I’m exporting in GRASS GIS 8.4.1 a large raster (25G file in fcell) with
``
export GDAL_CACHEMAX=2G
export GDAL_NUM_THREADS=2
r.out.gdal --o -f -c -m createopt=‘PREDICTOR=2,COMPRESS=DEFLATE,ZLEVEL=9,BIGTIFF=YES,NUM_THREADS=2,TILED=YES’ nodata=0 type=UInt32 input=input output
=/tmp/output.tif --overwrite
```
and is taking 5-6 hours producing a file of ~500 G.
I try to change the NUM_THREADS=2 and the compression but with no improvement.
Is there any tips that i can use to reduce the final geotif size and the time for exporting?

Thanks
Best
Giuseppe

2

There are probably bigger experts on this here, but based on my understanding PREDICTOR=3 might work better for floating point data. Depending on your use case, consider if you need it tiled, and as floating point data. Maybe this could also be helpful:

3

Thanks Anna!!!
Anyway, in my system (HPC) something strange is happening: compressed files are much larger than uncompressed ones. This happens for files larger than 4 GB, while for smaller files everything is fine.
In the end, I opted to use r.out.gdal without compression and then compress the files afterwards using gdal_translate.