How to find nearest road to single pair of latitude/longitude coordinates using R?

Question

I have a list of a couple of hundred latitude-longitude coordinate pairs for different locations. My goal is to obtain an estimate of the driving duration from a 'home' location to each of the coordinate pairs, using R.

I have had some success using the googleway package in R, but have (predictably) run into problems for locations which are far from a road, e.g. if the coordinates are for a mountain top. In these instances I would like to estimate instead the driving times to the nearest road to each problematic coordinate pair.

To illustrate, let's say my home location is;

home <- "Edinburgh, UK"

...and an example dataframe of locations I want to find driving times to could be;

location <- c("place_a", "place_b", "place_c") 
latitude <- c("56.87034", "57.69380", "57.36243")
longitude <- c("-4.199001", "-5.128715", "-5.104728")

df <- data.frame(location, latitude, longitude)

I can obtain distance/duration, etc., between home and place_a, and between home and place_b using something like;

(NB. You'll need your own Google Maps api key to replicate this section... )

library(googleway)
api_key <- [insert your Google Maps api key here!]

results <- google_distance(origins = home,
                  destinations = list(c("56.87034,-4.199001"),
                                      ("57.69380,-5.128715")),
                  mode = "driving",
                  key = api_key,
                  units = "imperial")

I get all the data I want using:

results$rows[[1]]

We run into trouble however when trying the same for the coordinates for place_c, which returns ZERO_RESULTS;

results2 <- google_distance(origins = home,
              destinations = ("57.36243,-5.104728"),
              mode = "driving",
              key = api_key,
              units = "imperial")

Here I think the issue is that the coordinates are for half-way up a mountain, so in this case I want to find the nearest road to the coordinates instead. I'd hoped to have some luck with the nearest_road function of googleway but cannot seem to get it working, e.g something like this doesn't work;

df_points <- read.table(text = "lat lon
                     57.36243 -5.104728", header = T)

nearest_road <- google_nearestRoads(df_points, key = api_key)

Can anyone advise what the problem is here? Or suggest a better solution altogether?!

Many thanks.

I can only guess that it's too far away from a road to find one. I had to move the coordiantes to `57.14743 -5.054728` for it to work. I couldn't see anything in [Google's documentation](https://developers.google.com/maps/documentation/roads/nearest) about how near to a road it has to be either. An alternative approach may be to find a shape file of roads in scotland and use geospatial operations, rather than Google's API. — SymbolixAU, Jan 09 '18 at 21:02
Thanks @SymbolixAU. I got there in the end thanks to a combination of the helpful post below and substituting in `googleway`s `google_distance` function to get past an error using `gmapsdistance`. — monkeytennis, Jan 11 '18 at 16:15

Neal Barsch · Accepted Answer · 2018-01-15T08:45:08.050

I am working on this very problem in a package that will be available on github soon (called spaceheater). In the meantime:

I would download the Open Street Maps shapefile from geofabrik for the country you are working with. For example Nigeria: http://download.geofabrik.de/africa/nigeria.html

EDITED WITH THE SUGGESTIONS OF monkeytennis(thanks!):

library(sp)
library(rgdal)
library(raster)
library(googleway)
library(geosphere)
library(foreach)
###I did Nigeria because I have it in my file downloaded, you would use UK###
roadshp <- readOGR(dsn="nigeria-latest-free.shp", 
layer="gis.osm_roads_free_1")
#Isolate primary roads (or secondary and tertiary) if you wish#
roads <- roadshp[roadshp$fclass %in% c("primary", "secondary", "tertiary"),]
#Use SpatialPoints for your gps coords
location <- c("place_a", "place_b", "place_c")
latitude <- c(8.641, 10.892, 11.797)
longitude <- c(6.0046, 11.146, 5.477)
df <- data.frame(location, latitude, longitude)
coordinates(df)=~longitude+latitude
sp1 <- SpatialPoints(df)
proj4string(sp1)=CRS("+proj=longlat +datum=WGS84 +no_defs +ellps=WGS84
                 +towgs84=0,0,0")
clodist <- dist2Line(sp1, roads)
df <- as.data.frame(df)
df$clodist <- clodist[,c("distance")]
df$lat <- clodist[,c("lat")]
df$lon <- clodist[,c("lon")]
iters <- nrow(df)
origin <- as.character("9.056, 7.497")
gc1 <- data.frame(round(df[,c("lat")],3), round(df[,c("lon")],3))
colnames(gc1) <- c("lat","lon")
df$lat <- as.character(gc1$lat)
df$lon <- as.character(gc1$lon)
gt2 <- paste(df[,c("lat")], df[,c("lon")], sep=",")
results <- google_distance(origins =origin, destinations= gt2,
            mode="driving",
            key="Your API Key Here")
results <-unlist(results)
results <- as.data.frame(results)
ttt <- head(results,-1)
ttt <- ttt[-c(iters+1), ]
m1 <- matrix(ttt, ncol=iters, byrow=TRUE)
distance <- as.data.frame(m1)
rownames(distance) <- c("Address", "DistanceKM", 
"DistanceM","TimeTextLow","TimeSecondsLow","TimeTextHigh","TimeSecondsHigh", 
"Status")

Thanks @Neal! This is a lot of effort and super helpful, much appreciated. My tired old laptop took all day completing the `clodist <- dist2Line(sp1, roads)` line but I got there in the end. One change I had to make to the above was to substitute `googleway`'s `google_distance` in for the `gmapsdistance` chunk of code. The original threw an error: `Error in rowXML[[dur]] : subscript out of bounds`. Seems like a common issue but I couldn't work out how to fix it. I'll keep an eye out for your spaceheater package. — monkeytennis, Jan 11 '18 at 16:06

How to find nearest road to single pair of latitude/longitude coordinates using R?

1 Answers1