使用NCO或Python测量多个netCDF文件的周平均值

sudo apt-get install nco

ncra -d time,0,-1,7 input*.nc output.nc

pip install netCDF4 numpy matplotlib

import netCDF4 as nc
import numpy as np

# 打开第一个文件，获取时间维度和变量信息
with nc.Dataset('input1.nc', 'r') as ncfile:
    time_var = ncfile.variables['time']
    time_units = time_var.units
    calendar = time_var.calendar
    var_names = [var for var in ncfile.variables if var!= 'time']

    # 计算总天数
    total_days = len(time_var)

    # 创建一个字典来存储每个变量的周平均值
    weekly_mean_dict = {var_name: np.zeros((total_days // 7,)) for var_name in var_names}

# 遍历所有文件
for i in range(1, 6):  # 假设有5个文件，文件名分别为input1.nc到input5.nc
    file_path = f'input{i}.nc'
    with nc.Dataset(file_path, 'r') as ncfile:
        for var_name in var_names:
            var_data = ncfile.variables[var_name][:]
            # 将变量数据按照周进行平均
            weekly_mean = np.mean(var_data.reshape(-1, 7), axis=1)
            weekly_mean_dict[var_name] += weekly_mean

# 将周平均值除以文件数量，得到最终的平均结果
for var_name in var_names:
    weekly_mean_dict[var_name] /= 5

# 创建输出的netCDF文件
with nc.Dataset('output.nc', 'w') as ncfile:
    # 创建时间维度
    time_dim = ncfile.createDimension('time', total_days // 7)
    time_var = ncfile.createVariable('time', 'f4', ('time',))
    time_var.units = time_units
    time_var.calendar = calendar

    # 将时间值设置为每周的中间值（这里只是简单示例，实际可能需要更准确的计算）
    time_var[:] = np.arange(3.5, total_days, 7)

    # 创建其他变量并写入周平均数据
    for var_name, weekly_mean in weekly_mean_dict.items():
        var = ncfile.createVariable(var_name, weekly_mean.dtype, ('time',))
        var[:] = weekly_mean