试图将Xarray WRF横截面保存到NetCDF时TypeError

Question

我试图将下面的Xarray DataArray保存为NetCDF，但是我会遇到错误“ ValueRorr：无法在变量'xy_loc';中推断dtype; xarray无法序列化任意的python对象”。数据是使用WRF软件包中的Vertcross生成的横截面。 I have resolved an earlier issue ("invalid value for attr") by removing the problem attributes as suggested here (

xarray.DataArray 'QVAPOR_cross' (XTIME: 2, vertical: 59, cross_line_idx: 239)
array([[[          nan,           nan,           nan, ...,
                   nan,           nan,           nan],
        [5.2142618e-03, 5.4577263e-03, 5.6551830e-03, ...,
         1.1576526e-02, 1.1481316e-02, 1.2421611e-02],
        [4.7921482e-03, 4.8334384e-03, 4.8439419e-03, ...,
         5.1765046e-03, 5.3730649e-03, 5.0648679e-03],
        ...,
        [3.1173588e-06, 3.1196325e-06, 3.1220573e-06, ...,
         1.7669336e-06, 1.8619023e-06, 1.9090951e-06],
        [3.2124219e-06, 3.2157229e-06, 3.2187045e-06, ...,
         1.9386096e-06, 1.9239062e-06, 1.8803454e-06],
        [          nan,           nan,           nan, ...,
         2.2164093e-06, 2.1845497e-06, 2.1732367e-06]],

       [[          nan,           nan,           nan, ...,
                   nan,           nan,           nan],
        [5.0503691e-03, 5.2436758e-03, 5.4539447e-03, ...,
         1.2035056e-02, 1.1335670e-02, 1.2419974e-02],
        [4.7799554e-03, 4.7886986e-03, 4.7935634e-03, ...,
         5.3098863e-03, 5.4153400e-03, 5.0202534e-03],
        ...,
        [3.1646039e-06, 3.1631198e-06, 3.1618749e-06, ...,
         1.8017681e-06, 1.8647409e-06, 1.8782921e-06],
        [3.2718322e-06, 3.2725636e-06, 3.2733096e-06, ...,
         1.8897241e-06, 1.8277435e-06, 1.7775615e-06],
        [          nan,           nan,           nan, ...,
         2.1244794e-06, 2.1224316e-06, 2.1411674e-06]]], dtype=float32)

Coordinates:
XTIME    (XTIME)          float64  720.0 780.0
Time     ()               object   None
xy_loc   (cross_line_idx) object   CoordPair(x=162.0, y=246.0, lat=...
vertical (vertical)       float32  970.83813 964.3068 ... 10.491312

Attributes:
FieldType : 104
description : Water vapor mixing ratio
units : kg kg-1
stagger :
coordinates : XLONG XLAT XTIME
projection : LambertConformal(stand_lon=-78.0, moad_cen_lat=38.0000114440918, truelat1=30.0, truelat2=60.0, pole_lat=90.0, pole_lon=0.0)
orientation : (162.0, 246.0) to (168.0, 8.0)
missing_value : 9.969209968386869e+36
_FillValue : 9.969209968386869e+36

Answer 1

错误消息说明了所有内容：“ ValueError：无法在变量'xy_loc'; xarray上推断dtype; xarray无法序列化任意python对象”。

xy_loc坐标包含一个类型对象的数组，这意味着它们不是均匀dtype的numpy数组，例如float64，int32 int32 < /code>等。相反，numpy数组将指针固定在Python对象上。第一个元素是称为coordpair（x = 162.0，y = 246.0，lat = ...。我不确定这是什么，看来Xarray也不是Xarray。

当Xarray写入NetCdf时，它序列化数据，这意味着它从python内存数据结构（例如dataarray，坐标等）到具有特定编码的文件。编码数据的协议

。 必须做一些预处理以准备数据以进行存储。

# add non-dimensional coordinates for each attribute of each CoordPair
da.coords['x'] = ('xy_loc', ), [c.x for c in xy_loc.values]
da.coords['y'] = ('xy_loc', ), [c.y for c in xy_loc.values]
da.coords['lat'] = ('xy_loc', ), [c.lat for c in xy_loc.values]
da.coords['lon'] = ('xy_loc', ), [c.lon for c in xy_loc.values]

# finally, replace the coord xy_loc itself with a positional index
da.corods['xy_loc'] = np.arange(len(da.xy_loc))

您 应该能够将对象保存为NetCDF