沿 2 个维度组合 2 个 Xarray DataArray（以便从粗网格获得更精细的网格）

Question

我有 2 个 DataArray，我必须将它们组合起来，但由于某些原因，这不起作用。（目的是创建一个具有更精细分辨率 (x2) 的网格）。

第一个数组 da_1 包含源数据：坐标对 z 的值 (x,y):

da_1 :

<xarray.DataArray (x: 3, y: 2)>
array([[1, 2],
       [3, 4],
       [5, 6]])
Coordinates:
  * x        (x) int64 0 1 2
  * y        (y) int64 8 9

然后，我创建了第二个数组 da_2：它具有与 da_1 相同的属性，但网格具有更精细的分辨率 (x2)，因此在 da_1 中x 坐标是 [0, 1, 2]，在 da_2 中是 [0, 0.5, 1, 1.5, 2]。对于 y 坐标：[8, 9] 变为 [8, 8.5, 9]。 z 值都是 NaN。

da_2 :

<xarray.DataArray (x: 5, y: 3)>
array([[nan, nan, nan],
       [nan, nan, nan],
       [nan, nan, nan],
       [nan, nan, nan],
       [nan, nan, nan]])
Coordinates:
  * x        (x) float64 0.0 0.5 1.0 1.5 2.0
  * y        (y) float64 8.0 8.5 9.0

最后，我必须替换 da_2 中 da_1 中存在的 NaN（具有相同的 (x,y) 坐标对）：(0,8), (0,9), (1,8), (1,9), (2,8) and (2,9).

预期结果：

<xarray.DataArray (x: 5, y: 3)>
array([[ 1., nan,  2.],
       [nan, nan, nan],
       [ 3., nan,  4.],
       [nan, nan, nan],
       [ 5., nan,  6.]])
Coordinates:
  * x        (x) float64 0.0 0.5 1.0 1.5 2.0
  * y        (y) float64 8.0 8.5 9.0

为此，我尝试将它们与 xarray.combine_by_coords() 结合使用，但失败了。

调用 combine_by_coords([da_1, da_2]) returns da_2 而不是预期的合并 DataArray，并且 combine_by_coords([da_2, da_1]) returns da_1.

我尝试了所有连接方法，但没有成功。

你知道如何得到预期的结果吗（da_2，值为da_1）？

可重现示例（这些数据的直观表示见下文）：

import xarray as xr

# Create first DataArray
da_1 = xr.DataArray([[1, 2], [3, 4], [5, 6]], dims=("x", "y"), 
        coords={"x": [0,1,2], "y": [8,9]})
print(da_1)
print("*"*50)

# Create second DataArray
nan = float("NaN")
da_2_data = [[nan, nan, nan],
            [nan, nan, nan],
            [nan, nan, nan],
            [nan, nan, nan],
            [nan, nan, nan]]

da_2 = xr.DataArray(da_2_data, dims=("x", "y"), 
        coords={"x": [0, 0.5, 1, 1.5, 2], "y": [8, 8.5, 9]})
print(da_2)
print("*"*50)

# Trying to combine
combined = xr.combine_by_coords([da_1, da_2])
print(combined)
print("*"*50)

expected_data = [[1, nan, 2],
                [nan, nan, nan],
                [3, nan, 4],
                [nan, nan, nan],
                [5, nan, 6]]

# Expected output (grid with resolution x2)
expected = xr.DataArray(expected_data, dims=("x", "y"), 
    coords={"x": [0, 0.5, 1, 1.5, 2], "y": [8, 8.5, 9]})
print(expected)
print("*"*50)

# If all is OK, we should get the same results as in da_1 for identical coordinates
x0_y8 = expected.sel(x=0, y=8).values
x0_y9 = expected.sel(x=0, y=9).values
x1_y8 = expected.sel(x=1, y=8).values
x1_y9 = expected.sel(x=1, y=9).values
x2_y8 = expected.sel(x=2, y=8).values
x2_y9 = expected.sel(x=2, y=9).values

assert(x0_y8 == 1)
assert(x0_y9 == 2)
assert(x1_y8 == 3)
assert(x1_y9 == 4)
assert(x2_y8 == 5)
assert(x2_y9 == 6)

Answer 1

一种解决方案是利用 xarray 和 pandas 之间的连接。您可以检查以下代码。如果你的数据非常大，比如气候科学中有数十亿行的数据框，那么唯一的问题就是速度。对于其他普通数据集，下面的方法应该没问题。

# import packages
import xarray as xr
import pandas as pd
import numpy as np

# construct your sample data
da_1 = xr.DataArray([[1, 2], [3, 4], [5, 6]], dims=("x", "y"), 
        coords={"x": [0,1,2], "y": [8,9]})

nan = float("NaN")
da_2_data = [[nan, nan, nan],
            [nan, nan, nan],
            [nan, nan, nan],
            [nan, nan, nan],
            [nan, nan, nan]]

da_2 = xr.DataArray(da_2_data, dims=("x", "y"), 
        coords={"x": [0, 0.5, 1, 1.5, 2], "y": [8, 8.5, 9]})

# build a function to convert xarray to pandas dataframe
def xr_to_df(input_xr):
    df = input_xr.to_dataframe()
    df = df.reset_index(drop=False)
    return df

# assign names to variables in "da_1" and "da_2"
# so you can combine them later
da_1 = da_1.rename("da_1")
da_2 = da_2.rename("da_2")

# conver both to pandas dataframes and combine the results
da_1_df = xr_to_df(da_1)
da_2_df = xr_to_df(da_2)

# now you can see that values from "da_1" and "da_2" are already matched on coordinates
da_df_combined = pd.merge(da_1_df,da_2_df,how='right')
print(da_df_combined)

# from now, conver the above dataframe back to xarray

# first get unique X and Y
# these should be natrually sorted from min to max
x = np.unique(da_df_combined['x'])
y = np.unique(da_df_combined['y'])

print("x:",x)
print("y:",y)

# then reshape the data to match the way it is structured
da_1_reshape =da_df_combined['da_1'].values.reshape(len(x),len(y))

# generate xarray and provide a name for the variable
# since you are only interested in values from "da_1", here we do "da_1" only
da_1_xr = xr.DataArray(da_1_reshape, coords=[('x', x),('y', y)])
da_1_xr = da_1_xr.rename("da_1")

# check your results
print(da_1_xr)

# use your way to doublecheck the values
x0_y8 = da_1_xr.sel(x=0, y=8).values
x0_y9 = da_1_xr.sel(x=0, y=9).values
x1_y8 = da_1_xr.sel(x=1, y=8).values
x1_y9 = da_1_xr.sel(x=1, y=9).values
x2_y8 = da_1_xr.sel(x=2, y=8).values
x2_y9 = da_1_xr.sel(x=2, y=9).values

assert(x0_y8 == 1)
assert(x0_y9 == 2)
assert(x1_y8 == 3)
assert(x1_y9 == 4)
assert(x2_y8 == 5)
assert(x2_y9 == 6)

沿 2 个维度组合 2 个 Xarray DataArray（以便从粗网格获得更精细的网格）

Combining 2 Xarray DataArrays along 2 dimensions (in order to obtain finer grid from coarse grid)

python

geospatial

python-xarray