to_crs creates different result from pyproj.transform

I’m not sure this is an issue for GeoPandas or pyproj, but parhaps due to always_xy option for pyproj.transform, to_crs creates different result from pyproj.transform.

Repro

Python and package versions

import sys

import pyproj
import geopandas

print(sys.version)
print(pyproj.__version__)
print(geopandas.__version__)

3.8.0 | packaged by conda-forge | (default, Nov 22 2019, 19:11:19) [Clang 9.0.0 (tags/RELEASE_900/final)]

2.4.2.post1

0.6.2

Reproject by pyproj.transform

x = -567637.0
y = -56204.0

# EPSG 4612: JGD2000, geographical, unit: degree
# https://epsg.io/4612
EPSG4612 = pyproj.Proj("EPSG:4612")
# RPSG 2443: JGD2000 / Japan Plane Rectangular CS I, projected, unit: meter
# https://epsg.io/2443
EPSG2443 = pyproj.Proj("EPSG:2443")

print(pyproj.transform(EPSG2443, EPSG4612, x, y))

(27.878013118155934, 128.92918943580798)

Here, 27.8780 is latitude and 128.929 is longitude.

Reproject by GeoDataFrame.to_crs

gdf = geopandas.GeoDataFrame(
    geometry = geopandas.points_from_xy(x = [x], y = [y]),
    crs = "EPSG:2443"
)
print(gdf.to_crs("EPSG:4612"))

geometry 0 POINT (123.47406 32.34878)

Swapping x and y produce correct result

gdf = geopandas.GeoDataFrame(
    geometry = geopandas.points_from_xy(x = [y], y = [x]),
    crs = "EPSG:2443"
)
print(gdf.to_crs("EPSG:4612"))

geometry 0 POINT (128.92919 27.87801)

Result produced by to_crs is similar for the one produced by transform with always_xy = True

print(pyproj.transform(EPSG2443, EPSG4612, x, y, always_xy=True))

(123.4740611549679, 32.34878175793676)

[snowman2]

The input CRS has the northing coordinate first.

>>> from pyproj import CRS
>>> CRS("EPSG:2443")
<Projected CRS: EPSG:2443>
Name: JGD2000 / Japan Plane Rectangular CS I
Axis Info [cartesian]:
- X[north]: Northing (metre)
- Y[east]: Easting (metre)
Area of Use:
- name: Japan - zone I
- bounds: (128.17, 26.96, 130.46, 34.74)
Coordinate Operation:
- name: Japan Plane Rectangular CS zone I
- method: Transverse Mercator
Datum: Japanese Geodetic Datum 2000
- Ellipsoid: GRS 1980
- Prime Meridian: Greenwich

And the output has the northing coordinate first.

>>> CRS("EPSG:4612")
<Geographic 2D CRS: EPSG:4612>
Name: JGD2000
Axis Info [ellipsoidal]:
- Lat[north]: Geodetic latitude (degree)
- Lon[east]: Geodetic longitude (degree)
Area of Use:
- name: Japan
- bounds: (122.38, 17.09, 157.65, 46.05)
Datum: Japanese Geodetic Datum 2000
- Ellipsoid: GRS 1980
- Prime Meridian: Greenwich

So, you need to input/output with north first and expect lat first:

>>> from pyproj import Transformer
>>> north = -56204.0
>>> east = -567637.0
>>> trans = Transformer.from_crs(2443, 4612)
>>> lat, lon = trans.transform(north, east)
>>> lat, lon
(32.34878175793677, 123.4740611549679)

However, geopandas uses always_xy, so it expects input x first and output x first always:

>>> transxy = Transformer.from_crs(2443, 4612, always_xy=True)
>>> lon, lat = transxy.transform(east, north)
>>> lon, lat
(123.4740611549679, 32.34878175793677)

Based on this, result_B seems to be correct.

[snowman2]

By “correct”, I mean after transformed the two points should be the same location on a map, or on the earth, right?

Yeah, you’re right. I originally skimmed your question 🙃 - bad habit of mine. I thought the order of the output was swapped as is shown in the example I provided. But, looking at it again, the values do differ. So only one of the two could have been correct.