[R-sig-Geo] polygonValues (raster): Very slow

[Nikhil Kaza]

This is not an R solution and I am not even sure if this speeds up  
your process. But in the past I have used starspan for this kind of  
work. It worked fairly well for me for large datasets. But it was a  
one off process that I didn't mind spending couple of hours. I also  
did a naive parallelization by breaking up the polygon files multiple  
parts and then assembling them back,  but if you really need the  
accurate proportion of cell area of a cell that falls across two  
polygons, this strategy wont work.

http://starspan.projects.atlas.ca.gov/doku/doku.php



Nikhil Kaza
Asst. Professor,
City and Regional Planning
University of North Carolina

nikhil.list at gmail.com

On Jun 30, 2010, at 10:12 AM, Agustin Lobo wrote:

> Hi!
> I'm trying:
>
>> eugrd025EFDC <- readOGR(dsn="eugrd025EFDC",layer="eugrd025EFDC")
> v <- polygonValues(p=eugrd025EFDC, Br, weights=TRUE)
>
> where
>
>> str(eugrd025EFDC,max.level=2)
> Formal class 'SpatialPolygonsDataFrame' [package "sp"] with 5 slots
>  ..@ data       :'data.frame':	18000 obs. of  5 variables:
>  ..@ polygons   :List of 18000
>  .. .. [list output truncated]
>  ..@ plotOrder  : int [1:18000] 17901 17900 17902 17903 17899 17898
> 17904 17897 17905 17906 ...
>  ..@ bbox       : num [1:2, 1:2] 2484331 1314148 6575852 4328780
>  .. ..- attr(*, "dimnames")=List of 2
>  ..@ proj4string:Formal class 'CRS' [package "sp"] with 1 slots
>
>> summary(Br)
> Cells:  13967442
> NAs  :  0
>
>
> Min.       0.00
> 1st Qu.    0.00
> Median     0.00
> Mean      48.82
> 3rd Qu.    0.00
> Max.    4999.00
>
> so quite large objects.
>
> The problem is that  polygonValues() has been running (and not
> completed the task) for
> more than 2 h on a intel core i7 machine with 16 Gb RAM (Dell
> Precision M6500), so a pretty powerful machine.
> Is there any way I could speed up this process?
> Also, is there anything I could do in order to take better advantage
> of the 8 processing threads?
> Currently, I see only 1 cpu working for R processes and the rest
> remain pretty inactive
>
> Thanks
>
> Agus
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo