1

I have a raster file and a SpatialLinesDataFrame in R. I need to find the distance between each pixel (at the center) and the nearest point in the line file. I believe this is the functional equivalent of the "near" function in ArcGIS. Is there a package or function that does this in R?

The raster is of a landscape with the year of deforestation as the cell value. The lines file is a series of roads- 34 total. I'm trying to determine the probability of deforestation around the roads using the distance of a point from the road as a variable.

The raster structure is:

# class       : RasterLayer 
# dimensions  : 9003, 7906, 71177718  (nrow, ncol, ncell)
# resolution  : 0.00025, 0.00025  (x, y)
# extent      : -52.7825, -50.806, -1.125, 1.12575  (xmin, xmax, ymin, ymax)
# coord. ref. : +proj=longlat +datum=WGS84 +no_defs +ellps=WGS84 +towgs84=0,0,0 
# data source : U:\Jacy\Research\TS_line_analysis\GIS_Project\NE_Hansen 
# names       : NE_Hansen 
# values      : 0, 14  (min, max)
Improve this question
2
  • What is the extent of your lines data? Are there any missing data values in the raster?
    – Spacedman
    Mar 30, 2017 at 6:55
  • @spacedman, the line extent is extent(ts_line) # class : Extent # xmin : -52.51523 # xmax : -51.07509 # ymin : -0.8724459 # ymax : 0.8546136 The raster file is a buffer around the lines (30km either way), so they overlap completely. There are a substantial number of NA values in my raster. I could give them a value without messing up my analysis if that will help. Thank you!
    – Jace
    Mar 30, 2017 at 18:28

1 Answer 1

Reset to default
5

Convert the raster to points, and use gDistance:

Some test data - some lines I had already and a raster over the lines:

> plot(lines)
> r = raster(extent(lines),100,100)

Convert to points:

> p = as(r,"SpatialPoints")

Use rgeos for distance calcs:

> require(rgeos)
> d = gDistance(p, lines, byid=TRUE)
Warning messages:
1: In RGEOSDistanceFunc(spgeom1, spgeom2, byid, "rgeos_distance") :
  Spatial object 2 is not projected; GEOS expects planar coordinates
2: In RGEOSDistanceFunc(spgeom1, spgeom2, byid, "rgeos_distance") :
  spgeom1 and spgeom2 have different proj4 strings

Ignore warning messages unless your data really has different projections. Resulting distances are in a matrix, because I have 3 lines and 100*100 points, I'm getting the nearest distance from the points to each of my three lines:

> dim(d)
[1]     3 10000

If you have a lot of lines then merge them down to a single feature.

Find the minimum point-to-line distance:

> dmin = apply(d,2,min)

Stick that in your raster:

> r[]=dmin
> plot(r)

Result, with lines on top:

enter image description here

Improve this answer
3
  • Thanks for the response! When I tried this, my dimensions for the distance matrix were different from the number of pixels in my raster. Any idea why this might be? 71177718 in the raster vs 10576571 in the output. Thank you!
    – Jace
    Mar 28, 2017 at 22:53
  • What dimensions are those? Is that the number of cells? Does it cause an error when you try and process it as I do? Has it clipped the output raster to the lines? I created my raster as en exact fit to the lines, but maybe your isn't like that. Suggest you edit your question and add some summary info about your raster and lines, especially the extents!
    – Spacedman
    Mar 29, 2017 at 13:57
  • hi @spacedman, I added some details about the raster structure and line file to the question. Does that help answer your question? When I used the code you suggested, the gDistance function ran and returned the same warnings, and the dmin/apply line also ran. I got an error when I tried to put the values back into the raster, because the number of cells did not match the number in the output of the distance function.
    – Jace
    Mar 29, 2017 at 19:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.