用于将数据数组导出到不工作的geotiff的降维

Question

我希望将xarray dataset ds的单个日期和变量ds导出到tif文件中。但是，我的数据数组有太多的维度，我无法理解如何有效地删除维度lsm_time。

    ds
    <xarray.Dataset>
    Dimensions:      (time: 2436, y: 28, x: 58, lsm_time: 1)
    Coordinates:
      * time         (time) datetime64[ns] 2015-05-02T12:00:00 ... 2021-12-31T12:...
        step         <U8 '12:00:00'
        surface      float64 0.0
      * y            (y) float64 36.01 36.11 36.21 36.31 ... 38.41 38.51 38.61 38.71
      * x            (x) float64 -7.64 -7.54 -7.44 -7.34 ... -2.24 -2.14 -2.04 -1.94
        spatial_ref  int32 0
    Dimensions without coordinates: lsm_time
    Data variables:
        ffmc         (time, y, x, lsm_time) float32 nan nan nan ... 88.93 88.53     
        dmc          (time, y, x, lsm_time) float32 nan nan nan ... 6.511 7.908 8.45
        dc           (time, y, x, lsm_time) float32 nan nan nan ... 406.2 428.5
        isi          (time, y, x, lsm_time) float32 nan nan nan ... 3.872 3.852
        bui          (time, y, x, lsm_time) float32 nan nan nan ... 15.08 16.11
        fwi          (time, y, x, lsm_time) float32 nan nan nan ... 5.303 5.486

导出数据数组会引发错误TooManyDimensions。

    ds.fwi.sel(time="07-14-2021").rio.to_raster("file_out.tif")

    raise TooManyDimensions(
    rioxarray.exceptions.TooManyDimensions: Only 2D and 3D data arrays supported. Data variable: fwi

在前面的步骤中，我已经删除了维度lsm_time，当我用陆地屏蔽lsm (单一日期)掩蔽dataset ds时，必须复制/添加dataset ds的时间维度。所以这可能是数据处理造成的错误。不过，我可以想办法不戴面具。

    lsm
    Dimensions:      (x: 58, y: 28, time: 1)
    Coordinates:
      * x            (x) float64 -7.64 -7.54 -7.44 -7.34 ... -2.24 -2.14 -2.04 -1.94
      * y            (y) float64 36.01 36.11 36.21 36.31 ... 38.41 38.51 38.61 38.71
      * time         (time) datetime64[ns] 2013-11-29
    Data variables:
        spatial_ref  int32 0
        lsm          (time, y, x) float64 0.0 0.0 0.0 0.0 0.0 ... 1.0 1.0 1.0 0.9996

    lsm = lsm.expand_dims({"new_time" : ds.time})
    lsm = lsm.rename({"time":"lsm_time"}).rename({"new_time" : "time"}).drop("lsm_time") #issue here: drop_dims removes variable..

    ds = ds.where(lsm>=threshold)

所以我已经申请了.drop("lsm_time")

但是，它仍然是

    print(ds.fwi.sel(time="07-14-2021").dims)
    > ('time', 'y', 'x', 'lsm_time')

当尝试.drop_dims时，它会删除所有数据变量。

    ds.drop_dims('lsm_time')
    <xarray.Dataset>
    Dimensions:      (time: 2436, y: 28, x: 58)
    Coordinates:
      * time         (time) datetime64[ns] 2015-05-02T12:00:00 ... 2021-12-31T12:...
        step         <U8 '12:00:00'
        surface      float64 0.0
      * y            (y) float64 36.01 36.11 36.21 36.31 ... 38.41 38.51 38.61 38.71
      * x            (x) float64 -7.64 -7.54 -7.44 -7.34 ... -2.24 -2.14 -2.04 -1.94
        spatial_ref  int32 0
    Data variables:
        *empty*

我错过了什么或者我做错了什么？谢谢你的帮助！

Answer 1

尺寸为1的尺寸可以使用.squeeze方法删除。

相反，可以使用.expand_dims添加大小为1的维度。

import xarray as xr
x = xr.tutorial.load_dataset("rasm")
y = x.isel(time=slice(0,1))

y.dims
# Frozen({'time': 1, 'y': 205, 'x': 275})

y.squeeze().dims
# Frozen({'y': 205, 'x': 275})

y.squeeze().expand_dims("newdim").dims
# Frozen({'y': 205, 'x': 275, 'newdim': 1})