Python计算气候数据温度距平时遇维度不兼容错误求助

作为Python初学者，我尝试基于AIRS 2010-2015年月度温度时序观测数据，参考ECMWF 1950-1980年月度气候态数据计算温度距平。二者网格分辨率不同，我已完成重格网处理，但执行距平计算时出现错误：

incompatible dimensions for a grouped binary operation: the group variable “month” is not a dimension on the other argument

可复现代码如下：

import pandas as pd
import numpy as np
from matplotlib import pyplot as plt
from mpl_toolkits.basemap import Basemap
import xarray as xr
import scipy

# import the target data
airs_temp_2010_15 = xr.open_dataset('temp_2010_15.nc')

# import the climatological data
climatology_ecmwf = xr.open_dataset('monthly_climatology_1950_1980.nc')

# essential to change the lon progression patten in compatible with airs target data
lon_name = 'longitude'  # whatever name is in the data

# Adjust lon values to make sure they are within (-180, 180)
climatology_ecmwf['_longitude_adjusted'] = xr.where(
    climatology_ecmwf[lon_name] > 180,
    climatology_ecmwf[lon_name] - 360,
    climatology_ecmwf[lon_name])

# reassign the new coords to as the main lon coords
# and sort DataArray using new coordinate values
climatology_ecmwf = (
    climatology_ecmwf
    .swap_dims({lon_name: '_longitude_adjusted'})
    .sel(**{'_longitude_adjusted': sorted(climatology_ecmwf._longitude_adjusted)})
    .drop(lon_name))
# finally rename the field again to longitude
climatology_ecmwf = climatology_ecmwf.rename({'_longitude_adjusted': lon_name})
print(climatology_ecmwf)

# let's change the grid resolution of the ecmwf climatology data
regrid_lon_axis_clim = np.arange(-180,180, 0.5)
regrid_lat_axis_clim = np.arange(-90,90, 0.5)

# apply the regridding over the climatology from ecmwf
regridded_clim_ecmwf = climatology_ecmwf.interp(longitude=regrid_lon_axis_clim, latitude=regrid_lat_axis_clim)
print(regridded_clim_ecmwf['t2m'].shape)

# regrid the target data as well to make sure the shapes are same for both the data
regrid_lon_axis = np.arange(-180, 180, 0.5)
regrid_lat_axis = np.arange(-90,90, 0.5)

# apply it over the airs temperature data
regridded_temp_2010_15 = airs_temp_2010_15.interp(Longitude=regrid_lon_axis,Latitude=regrid_lat_axis)
print(regridded_temp_2010_15['Temperature'].shape)

# Isolate one layer from target data which will be used as mask layer to mask all values of climatological data

regridded_temp_2010_15_1st_layer = regridded_temp_2010_15.isel(time=0)


# Mask the regridded climatology data at per regridded airs data
masked_regrid_clim_t2m = regridded_clim_ecmwf.where(regridded_temp_2010_15_1st_layer['Temperature'].values >0)
print(masked_regrid_clim_t2m)

# Now compute the anomaly
anom_2010_15 = regridded_temp_2010_15.groupby('time.month') - masked_regrid_clim_t2m

错误核心原因 #

当用groupby('time.month')对AIRS数据分组后，xarray要求气候态数据必须包含month维度，才能按月份匹配做减法。但你的气候态数据（masked_regrid_clim_t2m）是原始的长时序数据，没有单独的month维度，维度结构和分组后的AIRS数据完全不匹配。

修复步骤 #

1. 提取气候态月平均：先对ECMWF 1950-1980年的月度数据计算1-12月的多年平均，得到仅包含month维度的气候态场。
2. 统一变量名：将气候态数据的t2m重命名为Temperature，和AIRS数据的变量名保持一致，避免维度匹配时的隐性错误。
3. 对齐维度后计算距平：用分组后的AIRS数据减去对应月份的气候态均值。

修改后的关键代码如下：

# 1. 计算ECMWF气候态的月平均（得到1-12月的多年平均场）
clim_monthly_mean = regridded_clim_ecmwf.groupby('time.month').mean(dim='time')
# 2. 统一变量名，和AIRS数据的Temperature匹配
clim_monthly_mean = clim_monthly_mean.rename({'t2m': 'Temperature'})
# 3. 对气候态月平均场应用mask（和原逻辑保持一致）
masked_clim_monthly_mean = clim_monthly_mean.where(regridded_temp_2010_15_1st_layer['Temperature'] > 0)
# 4. 计算距平：分组后的AIRS数据自动匹配对应月份的气候态均值
anom_2010_15 = regridded_temp_2010_15.groupby('time.month') - masked_clim_monthly_mean

补充说明 #

• 原代码直接使用原始气候态时序数据，没有做月平均，导致维度和分组后的AIRS数据无法对齐。通过groupby('time.month').mean(dim='time')处理后，气候态数据维度变为(month, latitude, longitude)，正好和AIRS分组后的结构匹配。
• 变量名统一后，xarray会自动识别并对齐所有维度，无需手动指定匹配规则。

内容的提问来源于stack exchange，提问作者P Acharya