Virtual Warping

Rasterio has a WarpedVRT class that abstracts many of the details of raster warping by using an in-memory Warped VRT. A WarpedVRT can be the easiest solution for tiling large datasets.

For example, to virtually warp the RGB.byte.tif test dataset from its proper EPSG:32618 coordinate reference system to EPSG:3857 (Web Mercator) and extract pixels corresponding to its central zoom 9 tile, do the following.

from affine import Affine
import mercantile

import rasterio
from rasterio.enums import Resampling
from rasterio.vrt import WarpedVRT

with'tests/data/RGB.byte.tif') as src:
    with WarpedVRT(src, crs='EPSG:3857',
                   resampling=Resampling.bilinear) as vrt:

        # Determine the destination tile and its mercator bounds using
        # functions from the mercantile module.
        dst_tile = mercantile.tile(*vrt.lnglat(), 9)
        left, bottom, right, top = mercantile.xy_bounds(*dst_tile)

        # Determine the window to use in reading from the dataset.
        dst_window = vrt.window(left, bottom, right, top)

        # Read into a 3 x 512 x 512 array. Our output tile will be
        # 512 wide x 512 tall.
        data =, out_shape=(3, 512, 512))

        # Use the source's profile as a template for our output file.
        profile = vrt.profile
        profile['width'] = 512
        profile['height'] = 512
        profile['driver'] = 'GTiff'

        # We need determine the appropriate affine transformation matrix
        # for the dataset read window and then scale it by the dimensions
        # of the output array.
        dst_transform = vrt.window_transform(dst_window)
        scaling = Affine.scale(dst_window.height / 512,
                               dst_window.width / 512)
        dst_transform *= scaling
        profile['transform'] = dst_transform

        # Write the image tile to disk.
        with'/tmp/test-tile.tif', 'w', **profile) as dst:

Normalizing Data to a Consistent Grid

A WarpedVRT can be used to normalize a stack of images with differing projections, bounds, cell sizes, or dimensions against a regular grid in a defined bounding box.

The tests/data/RGB.byte.tif file is in UTM zone 18, so another file in a different CRS is required for demonstration. This command will create a new image with drastically different dimensions and cell size, and reproject to WGS84. As of this writing rio warp implements only a subset of gdalwarp’s features, so gdalwarp must be used to achieve the desired transform:

$ gdalwarp \
    -t_srs EPSG:4326 \
    -te_srs EPSG:32618 \
    -te 101985 2673031 339315 2801254 \
    -ts 200 250 \
    tests/data/RGB.byte.tif \

So, the attributes of these two images drastically differ:

$ rio info --shape tests/data/RGB.byte.tif
718 791
$ rio info --shape tests/data/WGS84-RGB.byte.tif
250 200
$ rio info --crs tests/data/RGB.byte.tif
$ rio info --crs tests/data/WGS84-RGB.byte.tif
$ rio bounds --bbox --geographic --precision 7 tests/data/RGB.byte.tif
[-78.95865, 23.5649912, -76.5749237, 25.5508738]
$ rio bounds --bbox --geographic --precision 7 tests/data/WGS84-RGB.byte.tif
[-78.9147773, 24.119606, -76.5963819, 25.3192311]

and this snippet demonstrates how to normalize data to consistent dimensions, CRS, and cell size within a pre-defined bounding box:

from __future__ import division

import os

import affine

import rasterio
from import CRS
from rasterio.enums import Resampling
from rasterio import shutil as rio_shutil
from rasterio.vrt import WarpedVRT

input_files = (
    # This file is in EPSG:32618
    # This file is in EPSG:4326

# Destination CRS is Web Mercator
dst_crs = CRS.from_epsg(3857)

# These coordinates are in Web Mercator
dst_bounds = -8744355, 2768114, -8559167, 2908677

# Output image dimensions
dst_height = dst_width = 100

# Output image transform
left, bottom, right, top = dst_bounds
xres = (right - left) / dst_width
yres = (top - bottom) / dst_height
dst_transform = affine.Affine(xres, 0.0, left,
                              0.0, -yres, top)

vrt_options = {
    'resampling': Resampling.cubic,
    'crs': dst_crs,
    'transform': dst_transform,
    'height': dst_height,
    'width': dst_width,

for path in input_files:

    with as src:

        with WarpedVRT(src, **vrt_options) as vrt:

            # At this point 'vrt' is a full dataset with dimensions,
            # CRS, and spatial extent matching 'vrt_options'.

            # Read all data into memory.
            data =

            # Process the dataset in chunks.  Likely not very efficient.
            for _, window in vrt.block_windows():
                data =

            # Dump the aligned data into a new file.  A VRT representing
            # this transformation can also be produced by switching
            # to the VRT driver.
            directory, name = os.path.split(path)
            outfile = os.path.join(directory, 'aligned-{}'.format(name))
            rio_shutil.copy(vrt, outfile, driver='GTiff')