[GRASS-dev] GRASS 7: r.stream.extract fails to read big input maps

Dear all,

I tried to extract a river network from a huge DEM:

rows: 180752

cols: 141312

cells: 25542426624

r.watershed (in an older revision of GRASS 7) had no problem with the DEM and created raster streams very nicely. However, I could not convert them to vector because r.to.vect complained the streams weren’t thinned properly. Since I would like to continue with the other r.streams.* addons anyway, I tried running r.stream.extract. On the entire DEM it failed with the error message below:

(Mon Apr 13 23:21:30 2015)

r.stream.extract --verbose elevation=dem_10m_nosefi_float@PERMANENT accumulation=dem_10m_nosefi_float_accum@Hydrologi threshold=1500 stream_length=5 memory=50000 stream_raster=dem_10m_nosefi_float_streams_ids stream_vector=dem_10m_nosefi_float_streams_ids

4.23% of data are kept in memory

Will need up to 1285.49 GB (1316340 MB) of disk space

Creating temporary files…

Loading input raster maps…

ERROR: Unable to load input raster map(s)

(Wed Apr 15 12:44:30 2015) Command finished (2243 min 0 sec)

From the error message (Unable to load input raster map(s)), I have no idea where I could start searching for the problem…

On a smaller catchment (with all in RAM calculation) r.stream.extract works fine…

Any ideas how to fix this? Running r.stream.extract per catchment (delineated earlier with r.watershed) would be a possible workaround…

Cheers

Stefan