使用Rasterio与Numpy创建TIFF文件失败，报错CPLE_OpenFailedError求助

Hey there! Let's break down the issues you're facing and fix them one by one.

First: Fix the "No Such File or Directory" Error #

The CPLE_OpenFailedError you're seeing isn't about your input file (since you can open montreal_30m.tif just fine) — it's about not being able to write the output file montreal_90m.tif. Here's why and how to fix it:

• Relative path confusion: You're using a relative filename, which means Rasterio tries to save the file in your current working directory. If that directory doesn't exist (or you don't have write permissions), it fails.
  • To check your current working directory, run this:

    import os
    print(os.getcwd())
    

  • Solution 1: Use an absolute path for the output file, like /home/your_username/data/montreal_90m.tif (adjust to your system's actual path).
  • Solution 2: If you want to save to a subfolder, make sure it exists first. Add this code before opening the output file:

    output_path = "montreal_90m.tif"
    # Create parent directories if they don't exist
    os.makedirs(os.path.dirname(output_path), exist_ok=True)
    

Second: Improve Your Transform Calculation #

While your hardcoded transform values might be correct, it's better to derive them from the original dataset instead of manually typing coordinates — this avoids typos and ensures accuracy. Replace your newtransform line with this:

# Calculate new transform using original dataset's parameters
new_transform = Affine(
    dataset.transform[0] * 3,  # Scale x cell size by 3
    dataset.transform[1],
    dataset.transform[2],  # Keep original top-left x coordinate
    dataset.transform[3],
    dataset.transform[4] * 3,  # Scale y cell size by 3
    dataset.transform[5]  # Keep original top-left y coordinate
)

Bonus: Simplify Resampling with Rasterio's Built-in Tools #

Your manual loop to downsample the raster works, but Rasterio has optimized resampling functions that are faster and cleaner. Here's a streamlined version of your code using Resampling.nearest (matches your current "take every 3rd pixel" logic):

import rasterio
from rasterio.plot import show
import numpy as np
from affine import Affine
from rasterio.enums import Resampling
import os

# Open input dataset
with rasterio.open('montreal_30m.tif') as dataset:
    # Calculate new dimensions (downscale by 3x)
    new_height = dataset.height // 3
    new_width = dataset.width // 3

    # Downsample using nearest neighbor (matches your current approach)
    new_band = dataset.read(
        1,
        out_shape=(new_height, new_width),
        resampling=Resampling.nearest
    )

    # Calculate new transform automatically
    new_transform = dataset.transform * dataset.transform.scale(
        (dataset.width / new_width),
        (dataset.height / new_height)
    )

    # Prepare output path and create directories if needed
    output_path = 'montreal_90m.tif'
    os.makedirs(os.path.dirname(output_path), exist_ok=True)

    # Write the new TIFF
    with rasterio.open(
        output_path,
        'w',
        driver='GTiff',
        height=new_height,
        width=new_width,
        count=1,
        dtype=np.float32,
        crs=dataset.crs,
        transform=new_transform
    ) as new_raster:
        new_raster.write(new_band, 1)

This code handles path creation, uses Rasterio's optimized resampling, and avoids hardcoded coordinates — making it more robust and easier to maintain.

内容的提问来源于stack exchange，提问作者Lisa Wang