Count lines crossing raster cells with R - r

Question

I have the following raster (shp file using readORG):
dput(summary_grid)
structure(list(class = structure("SpatialPolygonsDataFrame", package = "sp"),
bbox = structure(c(4346000, 3819000, 4445000, 3867000), .Dim = c(2L,
2L), .Dimnames = list(c("x", "y"), c("min", "max"))), is.projected = TRUE,
proj4string = "+proj=laea +lat_0=52 +lon_0=10 +x_0=4321000 +y_0=3210000 +ellps=GRS80 +units=m +no_defs",
data = structure(c("Min. :0 ", "1st Qu.:0 ", "Median :0 ",
"Mean :0 ", "3rd Qu.:0 ", "Max. :0 "), .Dim = c(6L,
1L), .Dimnames = list(c("", "", "", "", "", ""), " Id"), class = "table")), .Names = c("class",
"bbox", "is.projected", "proj4string", "data"), class = "summary.Spatial")
as well as the shp file containing lines:
dput(summary_lines)
structure(list(class = structure("SpatialLinesDataFrame", package = "sp"),
bbox = structure(c(4329488.96922647, 3429159.10800761, 4998503.48859431,
4055688.10547651), .Dim = c(2L, 2L), .Dimnames = list(c("x",
"y"), c("min", "max"))), is.projected = TRUE, proj4string = "+proj=laea +lat_0=52 +lon_0=10 +x_0=4321000 +y_0=3210000 +ellps=GRS80 +units=m +no_defs",
data = structure(c("Min. :0 ", "1st Qu.:0 ", "Median :0 ",
"Mean :0 ", "3rd Qu.:0 ", "Max. :0 ", "Min. : 448 ",
"1st Qu.:13229 ", "Median :28235 ", "Mean :27205 ",
"3rd Qu.:40724 ", "Max. :50608 ", "Min. : 485 ",
"1st Qu.:13731 ", "Median :29399 ", "Mean :28635 ",
"3rd Qu.:43159 ", "Max. :53607 ", "Min. :6519754 ",
"1st Qu.:8741416 ", "Median :9167928 ", "Mean :8894830 ",
"3rd Qu.:9414708 ", "Max. :9762259 ", "Fishing:121 ",
NA, NA, NA, NA, NA), .Dim = c(6L, 5L), .Dimnames = list(c("",
"", "", "", "", ""), c(" Id", " trip_id", " new_tr_id",
" species", " HELCOM_Gro")), class = "table")), .Names = c("class",
"bbox", "is.projected", "proj4string", "data"), class = "summary.Spatial")
In this shp file with the lines, each lines has a unique identification number called new_tr_id.
My goal is to create a raster file with the number if lines crossing each cell. A line (so the same new_tr_id) can cross the same cell X times, it would be counted X times. I added the figure below as example:
I have some interesting discussions using the package raster but I still can not find a proper answer to this issue.
Help is more than welcome :)

Answer 1

That is tricky. Counting is standard, and counting the total length is not that hard. But I think I come with something.
Example data
library(raster)
cds1 <- rbind(c(-180,-20), c(-140,55), c(10, 0), c(-140,-60))
cds2 <- rbind(c(-10,0), c(140,60), c(160,0), c(140,-55))
cds3 <- rbind(c(-125,0), c(0,60), c(40,5), c(15,-45))
lns <- spLines(cds1, cds2, cds3, crs="+proj=utm +zone=1 +datum=WGS84")
r <- raster(ncols=9, nrows=5, vals=1:45, crs="+proj=utm +zone=1 +datum=WGS84")
Standard count:
x <- rasterize(lns, r, fun='count')
We can break the line segments up by the raster cell borders (this will not work for very large rasters).
rsp <- rasterToPolygons(r)
lns2 <- intersect(lns, rsp)
r <- rasterize(lns2, r, fun='count')
But the result is not correct because if a segment touches a cell, it is counted as in.
So another approach would be to cut them and remove just a little at the edges. There are two parameters that you will have to play with for your data
wdth = .25
smallp = 100
b <- buffer(rsp, dissolve=FALSE, width=wdth)
j <- intersect(b, b)
j$area <- area(j)
small <- j[j$area < smallp, ]
small <- aggregate(small)
lns3 <- erase(lns, small)
r <- rasterize(lns3, r, fun='count')
plot(r)
lines(rsp)
plot(lns, col=rainbow(3), lwd=2, add=T)
Quite a work around, but the result looks good.