Mapping with leaflet

Interactive maps are easily made in R using the leaflet package. Let’s work through an example using the otter dataset. A leaflet map is created in layers, much like in ggplot2. The first layer calls the data using the leaflet() function. Then you add in basemap tiles, followed by multiple layers of markers or labels. In this case, we will use the default basemap with circle markers for each otter observation site.

otter<-read.csv("https://maddiebrown.github.io/ANTH630/data/sea_otter_counts_2017&2018_CLEANDATA.csv")
#see: https://www.htmlwidgets.org/showcase_leaflet.html
library(leaflet)
leaflet(otter) %>% 
  addTiles() %>%
  addCircleMarkers(lng=otter$longitude_E, lat=otter$latitude_N)

From the map above, we cannot tell whether there are any differences in the number of otters observed at each site. We can also create a custom color palette based on the number of otters observed per site. We use the colorNumeric() function to define a custom palette.

pal <- colorNumeric(palette="YlOrRd", domain=otter$n_otter)

This palette can then be assigned to the circle markers in the map.

leaflet(otter) %>% 
  addTiles() %>%
  addCircleMarkers(lng=otter$longitude_E, lat=otter$latitude_N, color = ~pal(n_otter))

Now we have colors, but because of overplotting it is difficult to differentiate between the sites. We can cluster the markers together using the clusterOptions argument.

leaflet(otter) %>% 
  addTiles() %>%
  addMarkers(clusterOptions = markerClusterOptions(),lng=otter$longitude_E, lat=otter$latitude_N)

This works for clustering the data and making it easier to assess where there are many or few observation sites. Hoewever, it still doesn’t solve our issue of conveying how many otters are observed at each site.

Try it

Many of the otter observation sites have only one or a few otters observed, while a few include large groups of otters. Perhaps we want to differentiate between the cases with only a few or many otters. We can manually set the different types of sites to unique aesthetic parameters. Using ~ifelse() manually create two classes of sites with different aesthetics based on the number of otters observed at each site.

Click for solution

library(leaflet)
leaflet(otter) %>% 
  addTiles() %>% addCircleMarkers(lng=otter$longitude_E,  lat=otter$latitude_N, radius = ~ifelse(n_otter > 5, 5, 1), color = ~ifelse(n_otter > 5, "red", "blue"))

Leaflet also allows us to make data points interactive either upon hovering or clicking.

Try it

Adding labels Using the otter sighting map we just created, add the site names as popup= information that appears upon clicking the points. Be sure to label the site names as site names, rather than just loading in the site names alone.

Click for solution

library(leaflet)
leaflet(otter) %>% 
  addTiles() %>% addCircleMarkers(lng=otter$longitude_E,  lat=otter$latitude_N, radius = ~ifelse(n_otter > 5, 5, 1), color = ~ifelse(n_otter > 5, "red", "blue"), popup= ~paste("Site name:", site_name, sep=" "))

Basemaps

There are a number of basemaps included in leaflet. Let’s look at a few of them here.

leaflet(otter) %>% 
 addProviderTiles(providers$Stamen.Toner) %>% addCircleMarkers(lng=otter$longitude_E,  lat=otter$latitude_N)
leaflet(otter) %>% 
 addProviderTiles(providers$CartoDB.Positron)%>% addCircleMarkers(lng=otter$longitude_E,  lat=otter$latitude_N)
leaflet(otter) %>% 
 addProviderTiles(providers$Stamen.Watercolor)%>% addCircleMarkers(lng=otter$longitude_E,  lat=otter$latitude_N)

Interactive graphics with plotly

One of the main libraries for making interactive plots in R is plotly. This library has extensive documentation and integration with ggplot.

Interactive boxplot

library(plotly)
reindeer <- read.csv("https://maddiebrown.github.io/ANTH630/data/Loring_et_al_reindeer_records.csv", stringsAsFactors = F)

fig1 <- plot_ly(reindeer, y= ~N.dogs, color=I("green"))
fig1 %>% add_boxplot(x=~Region)

Lollipop plot

library(plotly)
reindeer <- read.csv("https://maddiebrown.github.io/ANTH630/data/Loring_et_al_reindeer_records.csv", stringsAsFactors = F)
# transform data into a dataframe with counts per functional group
groupcount <- reindeer %>% group_by(Func.Grp) %>% summarise(count=n())

#make a basic ggplot
p <- ggplot(groupcount, aes(x=Func.Grp, y=count, text=Func.Grp)) + geom_point() + geom_segment(aes(x=Func.Grp, xend=Func.Grp, y=0, yend=count)) + theme_light() + ylab(paste("Number of Records")) + xlab("Functional Group") + ggtitle("Prevalence of functional groups in records of Native food surveys")

#make it interactive
p<- ggplotly(p)
p

Try it

The plot above looks nice but you’ll notice that the tooltip is not very useful. Make a new GrpExpl vector that lists each of the common names of the animals found in each functional group. One way to create this list involves going back to the original data and adding it at the time of summarizing.

Then, append this list and the number of records as lines in a tooltip box that appears upon hovering. Examine the ggplotly() help file if you are stuck.

Let’s also add the number of records to the y-axis as an automatically generated number that will update if the number of records changes.

While you’re at it, flip the coordinates to make the labels and relationships clearer.

Click for solution

groupcount<- reindeer %>% group_by(Func.Grp) %>% summarise(count=n(), GrpExpl=paste(unique(Common.name),collapse=", "))

p <- ggplot(groupcount, aes(x=Func.Grp, y=count, text=GrpExpl)) + geom_point() + geom_segment(aes(x=Func.Grp, xend=Func.Grp, y=0, yend=count)) + coord_flip() + theme_light() + ylab(paste("Number of Records (n= ", length(unique(reindeer$Record.ID)), ")", sep="")) + xlab("Functional Group") + ggtitle("Prevalence of functional groups in records of Native food surveys")

p<- ggplotly(p,tooltip=c("text","y"))
p

Tooltips are a helpful way to embed additional details into a plot without cluttering its appearance.

Specialized interactive graphs and widgets

Dygraph

It is possible to create dygraphs with an interactive time series using the dygraph package in R. Below is an example from html widgets and you are encouraged to learn more there if this type of graph is something you would like to explore.

# from https://www.htmlwidgets.org/showcase_dygraphs.html
library(dygraphs)
dygraph(nhtemp, main = "New Haven Temperatures") %>% 
  dyRangeSelector(dateWindow = c("1920-01-01", "1960-01-01"))

Sunburst plot

Sunburst plots are visually interesting plots that convey nested relationships and proportions. The package sunburstR can create these graphics but it does take quite a bit of data wrangling to get your data in the right format for this package to be able to understand your input.

Let’s make a 2-level sunburst plot using the Loring et al. data. In the code below, examine each object as it is created. This will reveal the final data structure that the package requires and how to transform your data to fit.

#devtools::install_github("timelyportfolio/sunburstR")
library(sunburstR)
library(tidyverse)
reindeer <- read.csv("https://maddiebrown.github.io/ANTH630/data/Loring_et_al_reindeer_records.csv", stringsAsFactors = F)

action <- reindeer %>%
  group_by(Record.ID) %>%
  filter(row_number() == 1) %>%
  summarize(event = paste(c(Func.Grp[!is.na(Func.Grp)],Food.type),collapse="-"))

sequences <- action %>%
  ungroup() %>%
  group_by(event) %>%
  summarize(count = n())

sequences$depth <- unlist(lapply(strsplit(sequences$event,"-"),length))

p  <- sequences %>%
  arrange(desc(depth), event) %>%
  sunburst(count = TRUE, 
  legend = FALSE, explanation = "function(d){return d.data.name}") 
p

Chord diagrams

We can also make chord diagrams denoting the relationship between variables.

#devtools::install_github("mattflor/chorddiag")
library(chorddiag)

# create a dummy variable
reindeer$count <- rep(1, nrow(reindeer))
#subset just the columns we want to use
reindeer2 <- reindeer[,c("Decade", "Func.Grp", "count")]
# pivot data into wide format
reindeer2<- reindeer2 %>% pivot_wider(names_from=Func.Grp, values_from = count, values_fn = sum )

#remove first column (decades)
reindeer3 <- reindeer2 %>% select(-Decade)
# set rownamees to decades
row.names(reindeer3) <- reindeer2$Decade
# convert to matrix
reindeer.mat <- as.matrix(reindeer3)

Try it

Make a palette with two distinct color ramps for each category of data. In this case, you’ll want distinct colors for each level of Decade and Func.Grp. UseRColorBrewer and consider the structure of the reindeer2 dataframe in making your color ramps.

Click for solution

#make a palette
library(RColorBrewer)
greenspal <- brewer.pal(9,"Greens")[2:9] 
bluespal <- brewer.pal(9, "Blues")[2:9]
blues2 <- colorRampPalette(bluespal)(length(unique(row.names(reindeer2))))
greens2 <- colorRampPalette(greenspal)(length(unique(colnames(reindeer2))))
allpal <- c(blues2,greens2)
# now we can create a chord diagram linking func.grp to decades by count.
chorddiag(reindeer.mat, type = "bipartite", showTicks = F, groupnameFontsize = 10, groupnamePadding = 10, margin = 150, groupColors = allpal)