2019_creating_maps_with_R.Rmd

---
title: "Creating (Web) Maps with R"
author: "Stephen Roecker"
date: "`r Sys.Date()`"
output:
  html_document:
    number_sections: yes
    toc: yes
    toc_float:
      collapsed: yes
      smooth_scroll: no
editor_options: 
  chunk_output_type: console
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(warning=FALSE, message=FALSE, cache=TRUE)
```


# Abstract

Most data has a spatial dimension to it. Knowing 'where' the data is coming from is often as crucial as knowing the 'what', 'when' and 'who' dimensions of a given dataset. Therefore it should be no surprise that R has a rich suite of packages for constructing maps and analyzing spatial data. R's capability has grown so much over the years that it's functionality rivals many dedicated geographic information systems (GIS). During this Meetup the basics for managing and mapping spatial data with be introduced, using the following packages: sf, ggplot2, tmap, mapview and leaflet.



# Example datasets

How you plot your data in R depends on what format it's in. R has several different formats for managing spatial data (e.g. sp vs sf), and different formats are only compatible with certain plotting systems. Thankfully converting between different spatial formats is not difficult.


## maps R package

```{r example-maps}

library(maps)

map(database = "state")
map(database = "state", region = "indiana")

```



## County Boundaries

```{r example-county}

library(USAboundaries)
library(sf)

cnty <- us_counties() 
cnty <- subset(cnty, !state_name %in% c("Alaska", "Hawaii"))

vars <- c("statefp", "countyfp")
plot(cnty[vars])

```



## USDA-NASS Corn Yield Data
https://github.com/potterzot/rnassqs

```{r example-nass, eval=FALSE}

devtools::install_github('potterzot/rnassqs')

library(rnassqs)

source("C:/Users/steph/Nextcloud/code/api_keys.R")

data(state)
st <- state.abb

corn_us <- lapply(st, function(x) {
  cat("getting", x, as.character(Sys.time()), "\n")
  tryCatch({
    corn = nassqs_yield(
      list("commodity_desc"="CORN",
           "agg_level_desc"="COUNTY",
           "state_alpha"=x
           ),
      key = nass_key
      )},
    error = function(err) {
      print(paste("Error occured:  ",err))
      return(NULL)
      }
    )
  })
corn_us <- do.call("rbind", corn_us)

save(corn_us, file = "C:/workspace2/corn_us.RData")
write.csv(corn_us, file = "nass_corn_us.csv", row.names = FALSE)

```

```{r yield2}

load(file = "C:/Users/Stephen.Roecker/Nextcloud/data/corn_us.RData")
corn_yield <- subset(corn_us, short_desc == "CORN, GRAIN - YIELD, MEASURED IN BU / ACRE")
corn_yield <- within(corn_yield, {
  Value      = as.numeric(Value)
  year       = as.numeric(year)
  state_name = NULL
  state      = state_alpha 
  })


cnty_corn <- merge(cnty, corn_yield, 
                   by.x = c("state_abbr", "countyfp"), 
                   by.y = c("state_alpha", "county_code"), 
                   all.x = TRUE
                   )

corn_states <- c("IL", "IA", "IN", "MI", "MN", "MO", "NE", "OH", "SD", "ND", "WI")

library(dplyr)
library(ggplot2)

group_by(corn_yield, state_alpha, year) %>%
  summarize(
    yield_low    = min(Value, na.rm = TRUE),
    yield_median = median(Value, na.rm = TRUE),
    yield_max    = max(Value, na.rm = TRUE)
    ) %>%
  filter(state_alpha %in% corn_states) %>%
  mutate(state = state_alpha,
         source = "NASS"
         ) %>%
  ggplot() +
  geom_line(aes(x = year, y = yield_median, col = source)) +
  geom_ribbon(aes(x = year, ymin = yield_low, ymax = yield_max), alpha = 0.25) +
  facet_wrap(~ state) +
  ylab("yield per county (bu/acre)") +
  ggtitle("USDA-NASS Corn Yields")
  # geom_point(data = yld_sum[yld_sum$state %in% corn_states, ], aes(x = 2018, y = yield_med, col = "NASIS"), size = 1) +
  # geom_ribbon(data = yld_sum2[yld_sum2$state %in% corn_states, ], aes(x = year, ymin = yield_low2, ymax = yield_max2, col = "NASIS"), alpha = 0.25) +
  # geom_pointrange(data = yld_sum[yld_sum$state %in% corn_states, ], aes(x = 2018, y = yield_med, ymin = yield_low2, ymax = yield_max2, col = source))

```



## Indiana General Soil Map

```{r example-statsgo}

IN <- sf::read_sf(dsn = "D:/geodata/soils/soils_GSMCLIP_mbr_2599033_03/wss_gsmsoil_IN_[2006-07-06]/spatial/gsmsoilmu_a_in.shp", layer = "gsmsoilmu_a_in")

# simplify polygons
IN <- rmapshaper::ms_simplify(IN)

```



## Kellogg Soil Survey Laboratory Data

```{r example-kssl}

library(aqp)
library(soilDB)

# download lab locations for the Miami soil series
miami <- fetchKSSL("Miami")
miami <- site(miami)
miami <- subset(miami, complete.cases(x, y))
miami <- within(miami, {
  lon = x
  lat = y
  })
head(miami)

```



# Construct and convert spatial objects

## point objects

```{r sp convert}

# construct

# sp object

library(sp)

miami_sp <- SpatialPointsDataFrame(
  data   = miami,
  coords = cbind(miami$lon, miami$lat),
  proj4string = CRS("+init=epsg:4326")
  )


#  sf object

library(sf)

miami_sf <- st_as_sf(
  miami,
  coords = c("lon", "lat"),
  crs = 4326
  )


# data structures
str(miami_sp, 2)
str(miami_sf)


# convert

miami_sf <- st_as_sf(miami_sp)
miami_sp <- as(miami_sf, "Spatial")

```



## Convert map objects

```{r map convert}

library(maptools)

st <- map("state", fill = TRUE, plot = FALSE)

# convert to sp object
st_sp <- map2SpatialPolygons(st, IDs = st$names)
proj4string(st_sp) <- CRS("+init=epsg:4326")
st_sp$state <- st$names


# loads data with ggplot2 package
st <- ggplot2::map_data("state")

```



## Convert with broom package

```{r broom}

library(broom)

st_tidy <- tidy(st_sp, region = "state") 
IN_tidy <- tidy(as(IN, "Spatial"), region = "MUSYM")

```




# Example maps with several R packages

## [ggplot2](https://ggplot2.tidyverse.org/)

ggplot2 plots data frames, but can also use sf objects (which are a special case of data frames). It can plot rasters, but only if they are converted to a data frame.


```{r ggplot2}

library(ggplot2)


# Lines
ggplot() +
  geom_point(data = miami, aes(x = lon, y = lat)) +
  geom_path(data = st, aes(x = long, y = lat, group = group)) +
  xlim(range(miami$lon)) +
  ylim(range(miami$lat)) +
  ggtitle("Location of Miami Lab Pedons")


# Polygons
ggplot() +
  geom_polygon(data = st_tidy, aes(x = long, y = lat, group = group, fill = id)) +
  coord_map(projection = "albers", lat0 = 39, lat1 = 45) +
  # remove legend
  guides(fill = FALSE)



# sf objects

# Polygons
ggplot() +
  geom_sf(data = cnty, aes(fill = statefp, lty = NA)) +
  geom_sf(data = miami_sf) +
  coord_sf(crs = "+init=epsg:5070") +
  guides(fill = FALSE)


# Facets

test <- subset(cnty_corn, year %in% 2012:2017)

ggplot() +
  geom_sf(data = test, aes(fill = Value, lty = NA)) +
  scale_fill_viridis_c(na.value = "transparent") +
  facet_wrap(~ year) +
  geom_path(data = st, aes(x = long, y = lat, group = group)) +
  ggtitle(corn_yield$short_desc[1])

```



## [ggmap](https://github.com/dkahle/ggmap)

ggmap expands ggplot2 to download and plot base maps.

```{r ggmap}

library(ggmap)

# build bound box and get base map via ggmap
bb <- sf::st_bbox(IN)
bb <- make_bbox(lon = bb[c(3, 1)], lat = bb[c(2, 4)])
gmap <- get_map(bb, maptype = "terrain", source = "osm")


# Lines
ggmap(gmap) +
  geom_path(data = IN_tidy, aes(x = long, y = lat, group = group))
  # geom_sf(data = IN, fill = NA, inherit.aes = FALSE) +
  # guides(fill = FALSE)

# geom_sf() doesn't work with ggmp, their is a systematic shift https://github.com/r-spatial/sf/issues/336

```



## [tmap](https://github.com/mtennekes/tmap)

The "t" in tmap stands for thematic, but tmap can also plot rasters natively. tmap's syntax is very similar to ggplot2, but with a few twists.

```{r tmap}

library(tmap)

tm_shape(IN) + tm_polygons("MUSYM", border.col = NULL) +
  tm_shape(cnty) + tm_borders() +
  tm_shape(miami_sf) + tm_dots() +
  tm_legend(legend.outside = TRUE)


# interactive web map 
tmap_mode("view")

tm_basemap("OpenStreetMap") +
  tm_shape(IN) + tm_borders()

```



## [mapview](https://r-spatial.github.io/mapview/)

```{r mapview}

library(mapview)

cols <- RColorBrewer::brewer.pal(50, "Paired")

test <- mapview(IN, zcol = "MUSYM", lwd = 0, col.regions = cols) +
  mapview(cnty, type = "l")

test

# export to html
mapshot(test, url = "C:/workspace2/test.html", selfcontained = FALSE)

```



## [leaflet](https://rstudio.github.io/leaflet/)

```{r leaflet}

library(leaflet)

test <- leaflet() %>%
  addProviderTiles("Esri.WorldImagery", group = "Imagery") %>%
  addPolygons(data = IN, fill = FALSE, color = "black", weight = 2)

test

# export to html
htmlwidgets::saveWidget(test, file = "C:/workspace2/test.html", selfcontained = FALSE)

```



# Additional Reading

Healy, K., 2018. Data Visualization: a practical introduction. Princeton University Press. [http://socviz.co/](http://socviz.co/)

Gimond, M., 2019. Intro to GIS and Spatial Analysis. [https://mgimond.github.io/Spatial/](https://mgimond.github.io/Spatial/)

Hijmans, R.J., 2019. Spatial Data Science with R. [https://rspatial.org/](https://rspatial.org/)

Lovelace, R., J. Nowosad, and J. Muenchow, 2019. Geocomputation with R. CRC Press. [https://bookdown.org/robinlovelace/geocompr/](https://bookdown.org/robinlovelace/geocompr/)

Pebesma, E., and R. Bivand, 2019. Spatial Data Science. [https://keen-swartz-3146c4.netlify.com/](https://keen-swartz-3146c4.netlify.com/)