Hello List, I need to reassign coordinates of a few points to have coordinates of the nearest raster cell. I KNOW that this is BAD practice but for the time being I assume that it does not affect my analysis (species distribution analysis). I think the datum of the point data is different than that of my raster data, but for now it is not important. I have already perused the "sp" and "raster" packages extensively. I guess I just need to be pointed in the right (WRONG) direction. I had a few thoughts with the distance function but I am now stuck. Maybe use "distance" then "which.min" then take the coordinates? Best Regards, Brian
Reassigning coordinates?
4 messages · Roger Bivand, Barry Rowlingson, Brian
On Tue, 2 Nov 2010, Brian Oney wrote:
Hello List, I need to reassign coordinates of a few points to have coordinates of the nearest raster cell. I KNOW that this is BAD practice but for the time being I assume that it does not affect my analysis (species distribution analysis). I think the datum of the point data is different than that of my raster data, but for now it is not important. I have already perused the "sp" and "raster" packages extensively. I guess I just need to be pointed in the right (WRONG) direction. I had a few thoughts with the distance function but I am now stuck. Maybe use "distance" then "which.min" then take the coordinates?
Maybe try nn2() in the RANN package with k=1 to find the nearest neighbour for each query point, with - I think - grid as data and points to move as query? Roger
Best Regards, Brian
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Roger Bivand Economic Geography Section, Department of Economics, Norwegian School of Economics and Business Administration, Helleveien 30, N-5045 Bergen, Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43 e-mail: Roger.Bivand at nhh.no
On Tue, Nov 2, 2010 at 10:00 AM, Brian Oney <zenlines at gmail.com> wrote:
Hello List, I need to reassign coordinates of a few points to have coordinates of the nearest raster cell. I KNOW that this is BAD practice but for the time being I ?assume that it does not affect my analysis (species distribution analysis). I think the datum of the point data is different than that of my raster data, but for now it is not important. I have already perused the "sp" and "raster" packages extensively. I guess I just need to be pointed in the right (WRONG) direction. I had a few thoughts with the distance function but I am now stuck. Maybe use "distance" then "which.min" then take the coordinates?
If you have some points in a SpatialPointsDataFrame and a grid as a
SpatialPixelsDataFrame then you can do an overlay. So for example,
using 'm' from help(SpatialPixelsDataFrame) and 'pts' as some random
clicking points:
> pts[1:10,]
coordinates zz
1 (179333, 330803) 0.76290682
2 (179600, 331026) 0.38275344
3 (180462, 331370) 0.75330357
...
> overlay(m,pts)
[1] 1930 1737 1434 1093 1115 1495 1705 2025 2146 2386 1780 728 487 1056 1256
[16] 1567 1778 1039 819 2465
- gives the index in 'm' where each point in pts is, hence something like:
cbind(coordinates(pts),coordinates(m)[overlay(m,pts),])
x y x y [1,] 179332.9 330802.6 179340 330820 [2,] 179599.7 331025.5 179580 331020 [3,] 180461.9 331369.9 180460 331380 [4,] 180533.7 331785.2 180540 331780 [5,] 180010.3 331734.6 180020 331740 gives the original and grid cell coordinates. something similar might work with package:raster objects... Barry
Hello Barry and Roger, Thank you for the suggestions. I got the nn2() function to work and took the coordinates of the nearest non-NA raster cell and assigned them to the points. I forgot to mention that I have raster cells that are non-NA and some that are not and my point lay in the NA zone (just barely). In this case, I set k=6 to get all surrounding raster cells (because some of the nearest cells still have an NA value) that selected the coordinate set from the raster cell whose nn distance to the next non-na cell was minimal. It was dirty but it worked. Thanks for the help, Brian # Code # rast.coords are a subset of relevant coodinates from the center of relevant raster cells. 'pc.na' is the xy-matrix of the coordinates needing reassigning. near <- data.frame(nn2(rast.coords, pc.nan, k=6)) na.near.cell1 <- near$nn.idx.2 reassign.coords <- xyFromCell(pc.clime, cell=na.near.cell) # Put them where they belong and voile.
On 11/2/2010 1:21 PM, Barry Rowlingson wrote:
On Tue, Nov 2, 2010 at 10:00 AM, Brian Oney<zenlines at gmail.com> wrote:
Hello List, I need to reassign coordinates of a few points to have coordinates of the nearest raster cell. I KNOW that this is BAD practice but for the time being I assume that it does not affect my analysis (species distribution analysis). I think the datum of the point data is different than that of my raster data, but for now it is not important. I have already perused the "sp" and "raster" packages extensively. I guess I just need to be pointed in the right (WRONG) direction. I had a few thoughts with the distance function but I am now stuck. Maybe use "distance" then "which.min" then take the coordinates?
If you have some points in a SpatialPointsDataFrame and a grid as a SpatialPixelsDataFrame then you can do an overlay. So for example, using 'm' from help(SpatialPixelsDataFrame) and 'pts' as some random clicking points:
> pts[1:10,]
coordinates zz 1 (179333, 330803) 0.76290682 2 (179600, 331026) 0.38275344 3 (180462, 331370) 0.75330357 ...
> overlay(m,pts)
[1] 1930 1737 1434 1093 1115 1495 1705 2025 2146 2386 1780 728 487 1056 1256 [16] 1567 1778 1039 819 2465 - gives the index in 'm' where each point in pts is, hence something like:
cbind(coordinates(pts),coordinates(m)[overlay(m,pts),])
x y x y [1,] 179332.9 330802.6 179340 330820 [2,] 179599.7 331025.5 179580 331020 [3,] 180461.9 331369.9 180460 331380 [4,] 180533.7 331785.2 180540 331780 [5,] 180010.3 331734.6 180020 331740 gives the original and grid cell coordinates. something similar might work with package:raster objects... Barry