Message-ID: <alpine.LFD.2.23.451.2008121836110.2892079@reclus.nhh.no>
Date: 2020-08-12T16:45:29Z
From: Roger Bivand
Subject: Mask from raster with less RAM
In-Reply-To: <CAL+ycWmwbpGiUNnBa6Kx7V1tTgF9DWwqa6wq0DTrvFqQ-N649Q@mail.gmail.com>
On Wed, 12 Aug 2020, Frederico Faleiro wrote:
> Hi guys, thank you for your reply.
>
> Jaime, I have tried, but I can't install rgdal needed to use raster and
> apparently other people too (
> https://stackoverflow.com/questions/57617895/how-to-install-rgdal-and-or-upload-raster-on-google-collaboration).
> Do you have a solution?
Use CRAN Windows or MacOS binaries, or install the system requirements.
Unless you can do that, for example because you do not control the
platform you are using (are you working on a cloud instance?), do not
install packages needing external software
from source.
> Hugo, I need the mask in vector format because the rasters have different
> resolutions, so I can't use raster as a mask. I have modified the reprex to
> be more precise about it (see below).
Please use GRASS. What you are trying to do is something that has "just
worked" in GRASS since it was first created. GRASS can be run from the
shell, from Python and from R (rgrass7). Or use SAGA, another fast raster
processor. Both mask from raster directly. Write a shell script for GRASS
to resample your rasters, mask them, and complete. You can also use GRASS
and SAGA from QGIS. You might use Python or R to make file name handling
"easier" than in a script. Neither GRASS nor SAGA use much memory unless
rasters are huge, and then they are lean.
Roger
> Steve, I think this approach has the same issue of Hugo Costa. I don't have
> the polygon of the mask, so I am trying to create one to apply in the other
> rasters of different resolutions.
>
> # reproducible example
> library(raster)
> # read data to create mask
> r <- raster(system.file("external/test.grd", package="raster"))
> r[!is.na(r)] <- 1
> pol <- rasterToPolygons(r, dissolve = T) # a lot of RAM to process the data
> # raster of different resolution
> res2 <- res(r) + 10
> r2 <- raster(extent(r), resolution = res2)
> r2[ ] <- 1
> # apply the mask
> r2.mask <- mask(r2, pol)
> # plot
> par(mfrow = c(1, 3))
> plot(r)
> plot(r2)
> plot(r2.mask)
>
> Cheers!
>
> Em ter., 11 de ago. de 2020 ?s 19:26, Stephen Stewart <
> stephen.stewart85 at gmail.com> escreveu:
>
>> Hi Frederico,
>>
>> It may not solve all of your RAM issues, but in this situation I would
>> skip the rasterToPolygons (which is also usually very slow) and use raster
>> math to propagate NAs.
>>
>> r <- raster(system.file("external/test.grd", package="raster"))
>> r[!is.na(r)] <- 1
>> # Can also be faster to do r = r / r, but add an offset (that cannot
>> result in 0) if you have valid 0s.
>> r2 <- raster(extent(r), res(r))
>> r2[ ] <- 1
>> r2.mask <- r * r2
>>
>> If you have a polygon to use as a mask, burn it in using the fasterize
>> package and then apply the above.
>>
>> Hope that helps.
>>
>> Cheers,
>>
>> Steve
>>
>> On Wed., 12 Aug. 2020, 6:43 am Frederico Faleiro, <fvfaleiro at gmail.com>
>> wrote:
>>
>>> Dear all,
>>>
>>> I would like to generate a mask from a raster, but my workflow needs a lot
>>> of RAM to process big rasters. I need this polygon mask to use in
>>> another's
>>> rasters.
>>> Do you know another approach that needs less RAM?
>>>
>>> # reproducible example
>>> library(raster)
>>> # read data to create mask
>>> r <- raster(system.file("external/test.grd", package="raster"))
>>> r[!is.na(r)] <- 1
>>> pol <- rasterToPolygons(r, dissolve = T) # a lot of RAM to process the
>>> data
>>> # apply the mask in another raster
>>> r2 <- raster(extent(r), res(r))
>>> r2[ ] <- 1
>>> r2.mask <- mask(r2, pol)
>>>
>>> Cheers!
>>>
>>> --
>>> Frederico Faleiro
>>> Postdoctoral Researcher in the INCT-EECBio (https://www.eecbio.ufg.br/)
>>> Department of Ecology | Institute of Biological Science | Federal
>>> University of Goi?s | Brazil
>>> RG: https://www.researchgate.net/profile/Frederico_Faleiro
>>> CV: http://lattes.cnpq.br/4926404840659003
>>>
>>> [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> R-sig-Geo mailing list
>>> R-sig-Geo at r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>
>>
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>
--
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: Roger.Bivand at nhh.no
https://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en