I am trying to manipulate some Brazilian Census data in R using the new "sf" package. I am able to import the data, but I get an error when I try to create the centroids of the original polygons
library(sf)
#Donwload data
filepath <- 'ftp://geoftp.ibge.gov.br/organizacao_do_territorio/malhas_territoriais/malhas_de_setores_censitarios__divisoes_intramunicipais/censo_2010/setores_censitarios_shp/ac/ac_setores_censitarios.zip'
download.file(filepath,'ac_setores_censitarios.zip')
unzip('ac_setores_censitarios.zip')
d <- st_read('12SEE250GC_SIR.shp',stringsAsFactors = F)
Now I try to create a new geometry column containing the centroid of column "geometry", but get an error:
d$centroid <- st_centroid(d$geometry)
Warning message:
In st_centroid.sfc(d$geometry) :
st_centroid does not give correct centroids for longitude/latitude data
How can I solve this?
All the GEOS functions underlying sf need projected coordinates to work properly, so you should run st_centroid on appropriately projected data. I don't know much about Brazil's available CRS's, but EPSG:29101 appears to work fine:
library(tidyverse)
d$centroids <- st_transform(d, 29101) %>%
st_centroid() %>%
# this is the crs from d, which has no EPSG code:
st_transform(., '+proj=longlat +ellps=GRS80 +no_defs') %>%
# since you want the centroids in a second geometry col:
st_geometry()
# check with
plot(st_geometry(d))
plot(d[, 'centroids'], add = T, col = 'red', pch = 19)
Related
I have the following polygon, defined using degrees latitude/longitude:
## Define latitude/longitude
lats <- c(64.25086, 64.24937, 63.24105, 63.22868)
lons <- c(-140.9985, -136.9171, -137.0050, -141.0260)
df <- data.frame(lon = lons, lat = lats)
polygon <- df %>%
## EPSG 3578; Yukon Albers projection
st_as_sf(coords = c('lon', 'lat'), crs = 3578) %>%
summarise(geometry = st_combine(geometry)) %>%
st_cast('POLYGON')
When I plot it on a map using Tmap, it appears in the Pacific Ocean off the coast of British Columbia, rather than in the middle of the Yukon:
library(sf)
library(sp)
library(tmap)
library(dplyr)
library(magrittr)
library(leaflet)
m <- tm_shape(data$study_boundary) + tm_borders(col = 'black',
lwd = 5,
zindex = 1000)
m
I am guessing that the problem is in using lat/long rather than UTMs because I have other polygons defined using UTMs that do appear where they (and the polygon defined above) are supposed to be. I found several other posts going the other way (UTM to lat/long) using spTransform, but I haven't been able to go lat/long to UTM with spTransform. I tried the code below:
poly_utm <- st_transform(polygon, crs = "+proj=utm+7")
But that didn't work either.
Thanks!
This (which I've improved by removing the pipe):
st_as_sf(df, coords = c('lon', 'lat'), crs = 3578)
creates a spatial points data frame using the numbers in the data frame for the coordinates, and the crs code of 3578 as the label for what those numbers represent. It does not change the numbers.
It looks like those numbers are actually lat-long coordinates, which means they are probable crs code 4326, the lat-long system used for GPS, also known as WGS 84. But it might not be. But probably is. Do check. So anyway, you should do:
df_unprojected = st_as_sf(df, coords = c('lon', 'lat'), crs = 4326)
df_projected = st_transform(df_unprojected, 3578)
The st_transform function does the actual change of the coordinate numbers and assigns the new CRS code to the spatial data metadata. That should give you a set of points you can then plot and check they are in the right place before you throw it into summarise and st_cast.
I'm trying to generate a raster and assign it a CRS projection. However, my CRS is in the new WKT format, and the raster() function is requiring me to provide a proj4string. Here's my code so far:
library(sf)
library(raster)
crs_dem <- st_crs(
'PROJCS["NAD_1983_2011_StatePlane_California_II_FIPS_0402",
GEOGCS["GCS_NAD_1983_2011",
DATUM["D_NAD_1983_2011",
SPHEROID["GRS_1980",6378137.0,298.257222101]],
PRIMEM["Greenwich",0.0],
UNIT["Degree",0.0174532925199433]],
PROJECTION["Lambert_Conformal_Conic"],
PARAMETER["False_Easting",2000000.0],
PARAMETER["False_Northing",500000.0],
PARAMETER["Central_Meridian",-122.0],
PARAMETER["Standard_Parallel_1",38.33333333333334],
PARAMETER["Standard_Parallel_2",39.83333333333334],
PARAMETER["Latitude_Of_Origin",37.66666666666666],
UNIT["Meter",1.0]]')
ext <- extent(1895000, 1935000, 579500, 616500)
grid <- raster(ext, resolution = c(40,40), crs = crs(dem))
The code above generates a raster with crs = NA. I've also tried assigning it with crs(grid)<- and projection(grid)<- with no luck. How can I get my specific CRS file to associate with this raster??
#slamballais's answer did the trick! I also found another (slightly less clean) method through trial and error last night, so here are both solutions.
Option 1:
test <- sp::CRS(crs_dem$input)
grid <- raster(ext, resolution = c(40,40), crs = test)
Option 2:
library(dplyr)
aoi <- ext %>%
as('SpatialPolygons') %>%
st_as_sf %>%
st_set_crs(crs_dem)
grid <- raster(ext, resolution = c(40,40), crs = projection(aoi))
raster expects text, so if you have a wkt format crs, you can use that directly. There is no need to create a more complex object.
crs_dem <- 'PROJCS["NAD_1983_2011_StatePlane_California_II_FIPS_0402",
GEOGCS["GCS_NAD_1983_2011", DATUM["D_NAD_1983_2011", SPHEROID["GRS_1980",6378137.0,298.257222101]],
PRIMEM["Greenwich",0.0], UNIT["Degree",0.0174532925199433]],
PROJECTION["Lambert_Conformal_Conic"], PARAMETER["False_Easting",2000000.0],
PARAMETER["False_Northing",500000.0], PARAMETER["Central_Meridian",-122.0],
PARAMETER["Standard_Parallel_1",38.33333333333334], PARAMETER["Standard_Parallel_2",39.83333333333334],
PARAMETER["Latitude_Of_Origin",37.66666666666666], UNIT["Meter",1.0]]'
library(raster)
r <- raster(crs=crs_dem)
or, if you start with an sf object
r <- raster(crs=crs_dem$input)
This is my first time doing any sort of spatial data visualization in R, and I'm stuck on a particular issue. I would like to clip a spatial polygon (specified by a series of lat/long coordinates) according to a world map, such that any part of the polygon which overlaps with a map polygon is removed. Using what I have in the below code as an example, I want to clip the rectangular spatial polygon so that only oceanic portions of the polygon remain.
I've found examples of how to retain the intersection between two spatial polygons, but I want to do the opposite. Perhaps there is a way to define the intersection, then "subtract" that from the polygon I wish to clip?
This might be a really basic question, but any tips will be appreciated!
Specify lat/long data:
x_coord <- c(25, 25, 275, 275)
y_coord <- c(20, -50, -50, 20)
xy.mat <- cbind(x_coord, y_coord)
xy.mat
Convert to spatial polygons object:
library(sp)
poly = Polygon(xy.mat)
polys = Polygons(list(poly),1)
spatial.polys = SpatialPolygons(list(polys))
proj4string(spatial.polys) = CRS("+proj=longlat +datum=WGS84 +no_defs
+ellps=WGS84 +towgs84=0,0,0")
Convert to spatial polygons data frame and export as shapefile:
df = data.frame(f=99.9)
spatial.polys.df = SpatialPolygonsDataFrame(spatial.polys, df)
spatial.polys.df
library(GISTools)
library(rgdal)
writeOGR(obj=spatial.polys.df, dsn="tmp", layer="polygon",
driver="ESRI Shapefile")
Plot world map and add .shp file:
map("world", wrap=c(0,360), resolution=0, ylim=c(-60,60))
map.axes()
shp <- readOGR("polygon.shp")
plot(shp, add=TRUE, col="blue", border=FALSE)
Here is a solution that stays in sf the entire time (I don't know sp), and illustrates constructing an sf object from scratch. st_difference create the geometry you want exactly, and then plotting can be done with the base plot method or the development version of ggplot which has geom_sf. I used map data from maps and rnaturalearth for this, you can adapt to your particular situation. Wrapping around the dateline is a little finicky regardless unfortunately.
library(tidyverse)
library(sf)
#> Linking to GEOS 3.6.1, GDAL 2.2.0, proj.4 4.9.3
library(rnaturalearth)
library(maps)
#>
#> Attaching package: 'maps'
#> The following object is masked from 'package:purrr':
#>
#> map
x_coord <- c(25, 25, 275, 275)
y_coord <- c(20, -50, -50, 20)
polygon <- cbind(x_coord, y_coord) %>%
st_linestring() %>%
st_cast("POLYGON") %>%
st_sfc(crs = 4326, check_ring_dir = TRUE) %>%
st_sf() %>%
st_wrap_dateline(options = c("WRAPDATELINE=YES", "DATELINEOFFSET=180"))
land <- rnaturalearth::ne_countries(returnclass = "sf") %>%
st_union()
ocean <- st_difference(polygon, land)
#> although coordinates are longitude/latitude, st_difference assumes that they are planar
plot(st_geometry(land))
plot(st_geometry(polygon), add = TRUE)
plot(st_geometry(ocean), add = TRUE, col = "blue")
ggplot() +
theme_bw() +
borders("world") +
geom_sf(data = ocean)
Created on 2018-03-13 by the reprex package (v0.2.0).
If I understand correctly what you want you can do it with the sf package using st_difference() and st_union()`.
Base on your code here is what you can do.
# world data
data("wrld_simpl", package = 'maptools')
# load sf package
library('sf')
# coerce sp object to sf
world <- st_as_sf(wrld_simpl)
rectangle <- st_as_sf(spatial.polys)
# difference between world polygons and the rectangle
difference <- st_difference(rectangle, st_union(world))
# coerce back to sp
difference <- as(difference, 'Spatial')
# plot the result
plot(difference)
I have a dataframe of points on map and an area of interest described as a polygon of points. I want to calculate the distance between each of the points to the polygon, ideally using the sf package.
library("tidyverse")
library("sf")
# area of interest
area <-
"POLYGON ((121863.900623145 486546.136633659, 121830.369032584 486624.24942906, 121742.202408334 486680.476675484, 121626.493982203 486692.384434804, 121415.359596921 486693.816446951, 121116.219703244 486773.748535465, 120965.69439283 486674.642759986, 121168.798757601 486495.217550029, 121542.879304342 486414.780364836, 121870.487595417 486512.71203006, 121863.900623145 486546.136633659))"
# convert to sf and project on a projected coord system
area <- st_as_sfc(area, crs = 7415L)
# points with long/lat coords
pnts <-
data.frame(
id = 1:3,
long = c(4.85558, 4.89904, 4.91073),
lat = c(52.39707, 52.36612, 52.36255)
)
# convert to sf with the same crs
pnts_sf <- st_as_sf(pnts, crs = 7415L, coords = c("long", "lat"))
# check if crs are equal
all.equal(st_crs(pnts_sf),st_crs(area))
I am wondering why the following approaches do not give me the correct answer.
1.Simply using the st_distance fun-doesn't work, wrong answer
st_distance(pnts_sf, area)
2.In a mutate call - all wrong answers
pnts_sf %>%
mutate(
distance = st_distance(area, by_element = TRUE),
distance2 = st_distance(area, by_element = FALSE),
distance3 = st_distance(geometry, area, by_element = TRUE)
)
However this approach seems to work and gives correct distances.
3.map over the long/lat - works correctly
pnts_geoms <-
map2(
pnts$long,
pnts$lat,
~ st_sfc(st_point(c(.x, .y)) , crs = 4326L)
) %>%
map(st_transform, crs = 7415L)
map_dbl(pnts_geoms, st_distance, y = area)
I'm new to spatial data and I'm trying to learn the sf package so I'm wondering what is going wrong here. As far as i can tell, the first 2 approaches somehow end up considering the points "as a whole" (one of the points is inside the area polygon so i guess that's why one of the wrong answers is 0). The third approach is considering a point at a time which is my intention.
Any ideas how can i get the mutate call to work as well?
I'm on R 3.4.1 with
> packageVersion("dplyr")
[1] ‘0.7.3’
> packageVersion("sf")
[1] ‘0.5.5’
So it turns out that the whole confusion was caused by a small silly oversight on my part. Here's the breakdown:
The points dataframe comes from a different source (!) than the area polygon.
Overseeing this I kept trying to set them to crs 7415 which is a legal but incorrect move and led eventually to the wrong answers.
The right approach is to convert them to sf objects in the crs they originate from, transform them to the one the area object is in and then proceed to compute the distances.
Putting it all together:
# this part was wrong, crs was supposed to be the one they were
# originally coded in
pnts_sf <- st_as_sf(pnts, crs = 4326L, coords = c("long", "lat"))
# then apply the transformation to another crs
pnts_sf <- st_transform(pnts_sf, crs = 7415L)
st_distance(pnts_sf, area)
--------------------------
Units: m
[,1]
[1,] 3998.5701
[2,] 0.0000
[3,] 751.8097
The sf package provides a great approach to working with geographic features, but I can't figure out a simple equivalent to the poly.counts function from GISTools package which desires sp objects.
poly.counts computes the number of points from a SpatialPointsDataFrame fall within the polygons of a SpatialPolygonsDataFrame and can be used as follows:
Data
## Libraries
library("GISTools")
library("tidyverse")
library("sf")
library("sp")
library("rgdal")
## Obtain shapefiles
download.file(url = "https://www2.census.gov/geo/tiger/TIGER2016/STATE/tl_2016_us_state.zip", destfile = "data-raw/states.zip")
unzip(zipfile = "data-raw/states.zip", exdir = "data-raw/states")
sf_us_states <- read_sf("data-raw/states")
## Our observations:
observations_tibble <- tribble(
~lat, ~long,
31.968599, -99.901813,
35.263266, -80.854385,
35.149534, -90.04898,
41.897547, -84.037166,
34.596759, -86.965563,
42.652579, -73.756232,
43.670406, -93.575858
)
Calculate points per polygon
I generate both my sp objects:
sp_us_states <- as(sf_us_states, "Spatial")
observations_spdf <- observations_tibble %>%
select(long, lat) %>% # SPDF want long, lat pairs
SpatialPointsDataFrame(coords = .,
data = .,
proj4string = sp_us_states#proj4string)
Now I can use poly.counts
points_in_states <-
poly.counts(pts = observations_spdf, polys = sp_us_states)
Add this into the sp object:
sp_us_states$points.in.state <- points_in_states
Now I've finished I'd convert back to sf objects and could visualise as follows:
library("leaflet")
updated_sf <- st_as_sf(sp_us_states)
updated_sf %>%
filter(points.in.state > 0) %>%
leaflet() %>%
addPolygons() %>%
addCircleMarkers(
data = observations_tibble
)
Question
Can I perform this operation without tedious conversion between sf and sp objects?
Try the following:
sf_obs = st_as_sf(observations_tibble, coords = c("long", "lat"),
crs = st_crs(sf_us_states))
lengths(st_covers(sf_us_states, sf_obs))
# check:
summary(points_in_states - lengths(st_covers(sf_us_states, sf_obs)))
st_covers returns a list with the indexes of points covered by each state; lengths returns the vector of the lenghts of these vectors, or the point count. The warnings you'll see indicate that although you have geographic coordinates, the underlying software assumes they are cartesian (which, for this case, will be most likely not problematic; move to projected coordinates if you want to get rid of it the proper way)