0_fuzz_site_coordinates.R

#!/usr/bin/env Rscript
library(tidyverse)

# "Fuzz" the true site coordinates to be used for the map
# This will only run if `data/site_coordinates_unfuzzed.csv` exists.
# That file is NOT included in the public GitHub repository.
# This script *is* included to help convince people that the published
# `data/site_coordinates.csv` function has a principled connection to
# reality...

## Haversine distance function to help sanity check fuzzing
haversine <- function(lat1, lon1, lat2, lon2) {
	lat1 <- lat1*pi / 180
	lon1 <- lon1*pi / 180
	lat2 <- lat2*pi / 180
	lon2 <- lon2*pi / 180
	delta_lat <- lat2 - lat1
	delta_lon <- lon2 - lon1
	x <- 1 - cos(delta_lat) + cos(lat1)*cos(lat2)*(1-cos(delta_lon))
	d <- 2*6371*asin(sqrt(0.5*x))
	return(d)
}

## Fuzz coordinates one at a time...
coords <- read_csv("data/site_coordinates_unfuzzed.csv")
for(i in 1:nrow(coords)) {
	while(TRUE) {
		lat <- coords$Latitude[i]
		lon <- coords$Longitude[i]
		# Pick a random distance between 100km and 200km and a random direction
		fuzz_dist_km <- rnorm(1, 150, 25)
		fuzz_bearing <- runif(1, 0, 2*pi)
		# km per deg lat/lon based on https://stackoverflow.com/a/1253545
		new_lat <- lat + (fuzz_dist_km / 110.574)*sin(fuzz_bearing)
		new_lon <- lon + (fuzz_dist_km / (111.320*cos(lat*pi/180)))*cos(fuzz_bearing)
		# Ensure distance between new and old coords is within +/- 1% of
		# intended fuzz distance, because who knows if above Stack
		# Overflow solution is correcgt?
		dist <- haversine(lat, lon, new_lat, new_lon)
		if(dist/fuzz_dist_km <= 1.01 & dist/fuzz_dist_km >= 0.99) {
			break
		}
	}
	coords$Latitude[i] <- new_lat
	coords$Longitude[i] <- new_lon
}
write_csv(coords, "data/site_coordinates.csv")