Transforming coordinates from one CRS to another with rasterio

Question

I'm trying to merge some rasters with rasterio.merge.merge. They are in the same CRS, EPSG:3857. I want the merged raster to cover the whole earth. Exactly the range (in coordinates): xmin, ymin, xmax, ymax = -180.0225, -90.0225, 179.9775, 90.0225.

The problem is that when using the bounds parameter of this function I have to give these limits in the same CRS of the rasters. How can I convert the tuple xmin, ymin, xmax, ymax from regular coordinates to projected coordinates in EPSG:3857?

Answer 1

This is how to achieve CRS transformation using rasterio (and optionally shapely):

import rasterio.warp
from rasterio.crs import CRS

# In GeoJSON format
x_min, y_min, x_max, y_max = bounds
feature = {
    "type": "Polygon",
    "coordinates": [
       [[x_max, y_min], 
        [x_max, y_max], 
        [x_min, y_max], 
        [x_min, y_min], 
        [x_max, y_min]]
   ]
}

# Project the feature to the desired CRS
feature_proj = rasterio.warp.transform_geom(
    CRS.from_epsg(4326),
    CRS.from_epsg(3857),
    feature
)

With the latest version of rasterio (tested with 1.2+) you can also use shapely, which simplify the code:

import rasterio.warp
from rasterio.crs import CRS
from shapely.geometry import box

# Project the feature to the desired CRS
feature_proj = rasterio.warp.transform_geom(
    CRS.from_epsg(4326),
    CRS.from_epsg(3857),
    box(*bounds)
)

Answer 2

simply transform your bb as 2 corners using pyproj:

from pyproj import CRS
from pyproj import Transformer

transformer = Transformer.from_crs(4326, 3857)

xmin, ymin = transformer.transform(-180.0225, -90.0225)
xmax, ymax = transformer.transform(179.9775, 90.0225)

Answer 3

For reasons discussed here and with the introduction of rasterio.warp.transform_bounds I think the way to do this would be:

from rasterio.warp import transform_bounds

bounds = (-180.0225, -90.0225, 179.9775, 90.0225)

xmin, ymin, xmax, ymax = transform_bounds(sar_img_crs.to_epsg(), 4326, *bounds)

Also, the answer by Pierrick above needs the always_xy=True flag to be set in the transformer.transform call for it to return values in the correct order.