Re-project match with a large raster

[JimShady]

I am trying to reproject a large raster, using a second raster as a template. This works fine on smaller rasters, but crashes with larger ones. I think I probably need to use window'd read and writes but I'm unsure how to implement this or if it's the correct solution. Thanks.

my_raster = rioxarray.open_rasterio("large.tif")

reference = rioxarray.open_rasterio("template_large_tif")

reprojected = my_raster.rio.reproject_match(reference)

reprojected.rio.to_raster("output.tif")

[snowman2]

Re-project match just uses the same CRS, transform, and shape of the other raster:

rioxarray/rioxarray/raster_array.py

Line 552 in 6b768ec

reprojected_data_array = self.reproject(

Examples referenced in solution:

• re-project a large raster
• reading files
• writing files

Solution:

import rasterio
import rasterio.vrt
import rioxarray

with rasterio.open("large.tif") as src, rasterio.open("template_large.tif") as template:
        with rasterio.vrt.WarpedVRT(
            src,
            crs=template.crs,
            transform=template.transform,
            height=template.height,
            width=template.width,
        ) as vrt:
            rds = rioxarray.open_rasterio(vrt, lock=False, parse_coordinates=False)
            rds.rio.to_raster("output.tif", tiled=True, windowed=True)

[JimShady]

Thanks for the help @snowman2 . Not very experienced with this library so struggling to implement your suggestion a bit. Looking at the URL you gave

env = rasterio.Env(
    GDAL_DISABLE_READDIR_ON_OPEN="EMPTY_DIR",
    CPL_VSIL_CURL_USE_HEAD=False,
    CPL_VSIL_CURL_ALLOWED_EXTENSIONS="TIF",
)
with env:
    with rasterio.open(url) as src: #this would be the file I want to reproject
        with rasterio.vrt.WarpedVRT(src, crs="EPSG:4326") as vrt: # I guess this is warping my file to 4326 (can I specify a template file here instead?)
            rds = rioxarray.open_rasterio(vrt) # This is opening the result I think. But I want the result written to file?

[JimShady]

i.e. I can't see from the example how I can use a 'template' raster, and/or how to write out the result.

[JimShady]

I used an AI bot to come up with this. Going to test it now.

output = "output.tif"

raster = rasterio.open("input_file.tif")
reference = rasterio.open("template.tif")

# Get the chunk size based on the raster dimensions
chunk_size = (raster.height // 10, raster.width // 10)

# Reproject the raster in smaller chunks
with rasterio.open(output , 'w', **raster.profile) as dst:
      dst.write(
          rasterio.warp.reproject(
              source=raster.read(),
              destination=dst,
              src_transform=raster.transform,
              src_crs=raster.crs,
              dst_transform=reference.transform,
              dst_crs=reference.crs,
              resampling=Resampling.bilinear,
              num_threads=1,
              chunk_size=chunk_size
          )
      )

[snowman2]

@JimShady I added a code example to reference.

[JimShady]

Thanks so much @snowman2 . Will give it a whirl over the weekend.

[RichardScottOZ]

reproject match is like this

template_raster = rioxarray.open_rasterio('template_raster.tif')

useraster = rioxarray.open_rasterio('useraster.tif')

finalraster = useraster.rio.reproject_match(template_raster)

finalraster = finalraster.assin_coords({"x":template_raster.x, "y":template_raster.y})   #so they match exactly

[JimShady]

Thanks @RichardScottOZ . Unfortunately I've still not managed to figure out how to do this with large rasters. I think I'm going to resort to 'raw' gdal on the command line via os.system() commands.