I read in the newspaper that the University of Hawaii was recruiting “citizen scientists” to help document the impact of this week’s “king tides” on our coastline. I uploaded some photos to the website but also found that we could download the dataset. I took this as an opportunity to learn something about geographic information systems (GIS) and R, and the end result was a pretty nice map of all the places that were photographed this week.
GIS in R
I knew nothing about GIS before starting this but there were some excellent articles on the web about it. One tutorial used the ggmap package. I just wanted to plot points on a map, and my prior experience with ggplot2 and this cheatsheet made it pretty easy.
The hardest part was all the installation requirements, specifically GDAL. There’s a lot of information on the web on how to install this, but it was a lot of googling that finally got it to work on Fedora. MacOS GDAL requires you to use packages available here.
Making a Map of Oahu
First I wanted a map of Oahu. This couldn’t be easier using ggmap. This produces a pretty watercolor looking map from Stamen.com.
library(tidyverse)
library(ggmap)
myLocation <- "Oahu"
myMap <- get_map(location = myLocation,
source = "stamen",
maptype = "watercolor",
crop = FALSE,
zoom = 10)
ggmap(myMap)
Then I wanted to put the points on the map. The dataset was in comma separated values format (albeit with this latin1 file encoding when I did this post first in Fedora). It was easy to load the dataset.
dat <- read.csv("../datasets/tides_20170526.csv",
# fileEncoding = "latin1",
colClasses = "character")
dat <- tbl_df(dat)
The file needed some cleaning for convenient use in the ggmap package.
dat
# Rename latitute and longitude columns for more convenient reference
names(dat)[9] <- "latitude"
names(dat)[10] <- "longitude"
# Make latitude and longitude numeric
dat$latitude <- as.numeric(dat$latitude)
dat$longitude <- as.numeric(dat$longitude)
# Convert date and time columns into POSIX format
dat$date_time <- paste(dat$Date, dat$Time)
dat$date_time <- strptime(dat$date_time, "%m/%d/%y %H:%M")
dat$date_time <- as.POSIXct(dat$date_time)
dat$LatLong <- paste0(dat$latitude, ":", dat$longitude)
I just wanted to see where people had been active in the last two days so I filtered the dataset by date.
minidat <- dat %>% filter(date_time > as.POSIXct("2017-05-25")) %>%
select(latitude, longitude, date_time, LatLong)
The resulting map was pretty sweet.
ggmap(myMap) + geom_point(aes(x = longitude, y = latitude), data = minidat, color = "black", alpha = 0.3)
I got virtually the same map using the qmap function. This function uses a Google Maps terrain map to create its base image.
qmap("Oahu") + geom_point(aes(x = longitude, y = latitude), data = minidat, color = "red", alpha = 0.3)
The resulting images from ggmap were pretty nice but I wanted more. Nowadays it’s customary to be able to pan and zoom like in Google Maps. The ggmap method produces a static map which is nice to look at, but it would be nice to see what points were taken on other islands. Or how about that blob of points on the south shore. Could we zoom in to see more readily where those points are?
googleVis version
I did some more research and found the gvisMap function from the googleVis package could provide a solution to this problem. Inspiration for how to do this came from this tutorial.
The special thing about gvisMap is that it requires you to feed it a variable containing latitude and longitude values in a specific latitude:longitude format. That’s why I did this above in the data cleaning steps.
It also allows you to write a label for each point. I just used the date and time label but if I were to spend a little more time on this, it would be simple to write a label that gave the record number, the time, and some brief description.
library(googleVis)
M1 <- gvisMap(minidat, "LatLong", tipvar = "date_time",
options=list(showTip=TRUE, showLine=F, enableScrollWheel=TRUE,
mapType='satellite', useMapTypeControl=TRUE, width=800,height=800))
The next command should be plot(M1) but this launches a javascript based map running locally. I have to figure out how to integrate this into a Jekyll blog like this one, but for the time being, you can see the result here. I copied the code from running print(M1, 'chart') onto an html page and voila, the map works just like it should!
Discussion
It wasn’t too hard to plot points on a map. If I’m going to do more of this, I think the googleVis is probably better for dynamic presentation than ggmap but the ggmap might be nicer for static presentation.
Looking at the actual results, it comes as no surprise that urban Honolulu is well documented. There are lots of people living here and so Waikiki and Hawaii Kai were covered pretty well. Rural areas were much less well documented, so that’s something for the project to work on!
Conclusions
GIS in R is feasible for a relative newbie to GIS. Citizen scientists on the King Tides Project documented the urban Honolulu area better than rural areas.