如何用Python读取存储在多个文件夹中的多个NetCDF文件？

Got it, let's tackle this problem! The core idea is to first collect all the NetCDF file paths across your target folders, then feed that list to xarray.open_mfdataset(). Here are three reliable, easy-to-use approaches:

方法1：用glob递归匹配（Python 3.5+） #

The glob module's recursive mode lets you match files in all subfolders with a simple pattern. This is the quickest option for most cases:

import glob
import xarray as xr

# 匹配根目录下所有子文件夹中的.nc文件
file_paths = glob.glob('F:/netcdf/example/**/*.nc', recursive=True)

# 批量合并读取
dsmerged = xr.open_mfdataset(file_paths)

Tip: If you only want to target specific subfolders (not all), adjust the pattern—for example, 'F:/netcdf/example/year*/month/*.nc' to match files in year folders nested under the example directory.

方法2：手动遍历文件夹（高度灵活） #

Use os.walk to traverse every directory and collect files manually. This is great if you need to add custom filters (like excluding certain folders or matching specific filename patterns):

import os
import xarray as xr

root_dir = 'F:/netcdf/example'
file_paths = []

# 遍历根目录下的所有文件和子目录
for dirpath, _, filenames in os.walk(root_dir):
    for filename in filenames:
        # 只收集.nc文件，可添加额外条件（比如文件名包含"temp"）
        if filename.endswith('.nc'):
            file_paths.append(os.path.join(dirpath, filename))

# 合并读取
dsmerged = xr.open_mfdataset(file_paths)

方法3：用pathlib（现代路径处理） #

The pathlib module offers a more object-oriented way to handle file paths, which is cross-platform and intuitive:

from pathlib import Path
import xarray as xr

root_path = Path('F:/netcdf/example')

# 递归查找所有.nc文件并转换为字符串列表
file_paths = [str(path) for path in root_path.rglob('*.nc')]

# 合并读取
dsmerged = xr.open_mfdataset(file_paths)

额外注意事项 #

• Ensure all your NetCDF files have compatible structures (same variables, dimensions, etc.). If not, you can use parameters like combine='nested' or concat_dim='time' to control how xarray merges them.
• For large numbers of files, enable parallel processing (requires dask) with xr.open_mfdataset(file_paths, parallel=True) to speed up reading.

内容的提问来源于stack exchange，提问作者Vishal Singh