2

I need to take the locations (Lat/Lon) in DF1 (3000 locations) and find the Haversine distance to the nearest location in DF2 (1500 locations) using R.

I can get this to work by hardcoding locations from DF2 or creating a column for each item in DF2 which is very cumbersome.

Any ideas on how to accomplish this efficiently?

Data structures
DF1
Lat   Lon

DF2
Lat   Lon
RROBINSON
  • 191
  • 1
  • 2
  • 11
  • this question has been answered previously although it is hard to find. Here is the link: https://stackoverflow.com/questions/40074926/merge-by-lat-lon-in-r, try the one using `geosphere` (there is only one answer now, but that may change one day) – sconfluentus Jan 10 '20 at 05:35

1 Answers1

1

Try this:

# load packages
install.packages("tidyverse")
library(geosphere)
install.packages("rjson")
library(rjson)

# make a function to retrieve test data


get_latlon <- function(x){ 
  url <- paste0("https://api3.geo.admin.ch/rest/services/api/SearchServer?searchText=",paste(x),"&type=locations")
  result <- rjson::fromJSON(file = URLencode(url))


  as_tibble(result$results[[1]]) %>% 
    mutate(attr_names = names(attrs)) %>% 
    spread(attr_names, attrs) %>% 
    unnest(cols = c(detail, featureId, geom_quadindex, geom_st_box2d, label, lat, 
                    lon, num, origin, rank, x, y, zoomlevel)) %>% 
    select(detail,lat,lon)
  }


# retrieve test data

cities1 <- c("spiez","zumikon","winterthur","neuenburg")
cities2 <- c("zurich","bern","lausanne")

cities1 %>% map(get_latlon) %>% bind_rows() -> DF1
cities2 %>% map(get_latlon) %>% bind_rows() -> DF2


# make a combined dataframe 
names(DF1) <- paste0(names(DF1), ".a")
names(DF2) <- paste0(names(DF2), ".b")
crossing(DF1,DF2) -> data

# function to calculate the Harversine distance

haversine <- function(lon1, lat1, lon2, lat2, r = 6378137) {

  if(!is.numeric(c(lon1, lat1, lon2, lat2)))
    stop("Inpust must be numeric")

  # Convert degrees to radians
  lon1 <- lon1 * pi / 180
  lat1 <- lat1 * pi / 180
  lon2 <- lon2 * pi / 180
  lat2 <- lat2 * pi / 180

  delta.lon <- (lon2 - lon1)
  delta.lat <- (lat2 - lat1)
  a <- sin(delta.lat/2)^2 + cos(lat1) * cos(lat2) *
    sin(delta.lon/2)^2
  c <- 2 * asin(min(1,sqrt(a)))
  distance <- r * c

  return(distance) # Distance
}


# find the smallest distances for locations in DF1
data %>%
  group_by(detail.a,detail.b) %>% 
  mutate(haversine=haversine(lat.a,lon.a,lat.b,lon.b)) %>% 
  group_by(detail.a) %>% 
  slice(which.min(haversine))
captcoma
  • 1,768
  • 13
  • 29