Reprojecting CRS from Geocode API fails to match shapefile CRS

Question

I have a shapefile obtained from this link.

library(rgdal)
library(rgeos)

dir <- tempdir()
unzip('~/Downloads/4106902.zip', exdir = paste0(dir, '/4106902'))

shape <- readOGR(paste0(dir, '/4106902'), '4106902')
shape <- gBuffer(shape, byid = T, width = 0) #correcting orphaned hole problem

When I check the CRS associated with the file I get:

> proj4string(shape)
[1] "+proj=utm +zone=22 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0"

This is equivalent to the EPSG:32722, according to spatialreference.org.
I need to locate a point obtained from Google Geocode API in this shapefile, so I'll read the latlong values and reproject them to the EPSG:32722:

google <- CRS("+init=epsg:4326")
point <- SpatialPoints(cbind(-49.25848, -25.45543),   
                       proj4string = google)
point.rep <- spTransform(point, CRS('+init=epsg:32722')) 

> summary(point.rep)
Object of class SpatialPoints
Coordinates:
          min       max
lng  675099.7  675099.7
lat 7183477.2 7183477.2
Is projected: TRUE 
proj4string :
[+init=epsg:32722 +proj=utm +zone=22 +south +datum=WGS84 +units=m +no_defs
+ellps=WGS84 +towgs84=0,0,0]
Number of points: 1

However, when I try to match the locations, I receive an error message stating that the CRS are not identical.

> over(point.rep, shape)
Error in .local(x, y, returnList, fn, ...) : 
  identicalCRS(x, y) is not TRUE

If I try to reproject the coordinates from google using the CRS extracted directly from the shapefile, instead of using the equivalent EPSG code, I get wrong lat value from the reprojection:

point.rep2 <- spTransform(point, proj4string(shape))

> summary(point.rep2)
Object of class SpatialPoints
Coordinates:
           min        max
lng   675099.7   675099.7
lat -2816522.8 -2816522.8
Is projected: TRUE 
proj4string :
[+proj=utm +zone=22 +datum=WGS84 +units=m +no_defs +ellps=WGS84
+towgs84=0,0,0]
Number of points: 1

> over(point.rep2, shape)
   ID_ NAME1_ NAME2_ PARTS_ POINTS_ LENGTH_ AREA_
1 <NA>   <NA>   <NA>     NA      NA      NA    NA

My current workaround is to force the CRS of my points after the reprojection, allowing me to find the correct location of the point in my shapefile:

> proj4string(point.rep) <- proj4string(shape) 
Warning message:
In ReplProj4string(obj, CRS(value)) :
  A new CRS was assigned to an object with an existing CRS:
+init=epsg:32722 +proj=utm +zone=22 +south +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0
without reprojecting.
For reprojection, use function spTransform

> over(point.rep, shape)
              ID_ NAME1_ NAME2_ PARTS_ POINTS_  LENGTH_    AREA_
1 410690205010318   <NA>   <NA>      1      27 1.661884 0.115704

Am I missing something? Why do I fail do reproject my point using the CRS from the shapefile?

Answer 1

Indeed, you have objects with different projection strings.

> identicalCRS(shape, point.rep)
[1] FALSE

Inspecting them makes it clearer:

> shape@proj4string
CRS arguments:
 +proj=utm +zone=22 +datum=WGS84 +units=m +no_defs +ellps=WGS84 +towgs84=0,0,0
> point.rep@proj4string
CRS arguments:
 +init=epsg:32722 +proj=utm +zone=22 +south +datum=WGS84 +units=m +no_defs +ellps=WGS84
+towgs84=0,0,0 

you will note that after the reprojection they do not have the same identical CRS string. Reprojecting the two of them to CRS("+proj=longlat +ellps=WGS84 +datum=WGS84") worked out for me:

> crs.geo <- CRS("+proj=longlat +ellps=WGS84 +datum=WGS84")
> shape <- spTransform(shape, crs.geo)
> point.rep <- spTransform(point, crs.geo)
> identicalCRS(shape, point.rep)
[1] TRUE

and so your over() command should work out. I noted though that the coordinates you create are out of your shape's bbox, so it finds no overlaying between the two objects.