python-xarray从一个DataArray复制掩码到另一个 [英] python-xarray copy mask from one DataArray to another

问题描述

我得到了一个简单的案例:

I got this to work for a simple case:

arr2 = xr.DataArray((np.arange(16)-8).reshape(4, 4), dims=['x', 'y'])
arr3 = xr.DataArray(np.arange(16).reshape(4, 4), dims=['x', 'y'])
<xarray.DataArray (x: 4, y: 4)>
array([[ nan,  nan,  nan,  nan],
[ nan,  nan,  nan,  nan],
[ nan,   9.,  10.,  11.],
[ 12.,  13.,  14.,  15.]])
Dimensions without coordinates: x, y

但是，我在申请NetCDF文件时遇到了麻烦. 我有两个数据集:重要的波高(Hs)和风速(ws). 我想使用Hs ＜0的掩码并将其应用于ws. 数据集的大小为[time = 1，lat = 81，lon = 131].在未来的某个时候，我的ws DataArray的大小会略有不同，例如[time = 1，ens = 10，lat = 81，lon = 131].

However, i'm having troubling applying to NetCDF files. I have two datasets: significant wave height (Hs) and wind speed (ws). I would like to use the mask of where Hs<0 and apply it to ws. The size of the datasets are [time=1,lat=81,lon=131]. There will be a time in the futures where my ws DataArray will be a slightly different size e.g. [time=1,ens=10,lat=81,lon=131].

如果我尝试:

f = xr.open_dataset('CCSM4_ens1_19821201_19831130_ws10_0_NAtl_DJFmean.nc')
ws10 = f.ws10
f = xr.open_dataset('ww3.Hs.19820901_19830831_NAtl_DJFmean.nc')
hs = f.hs
ws10_masked = ws10.where(hs > 0)

ws10_masked看起来像:

ws10_masked looks like:

xarray.DataArray (time: 1, lat: 81, lon: 131, latitude: 81, longitude: 131)
array([[[[[ nan, ...,  nan],
      ..., 
      [ nan, ...,  nan]],
     ..., 
     [[ nan, ...,  nan],
      ..., 
      [ nan, ...,  nan]]],
      ..., 
      [[[ nan, ...,  nan],
      ..., 
      [ nan, ...,  nan]],
      ..., 
      [[ nan, ...,  nan],
      ..., 
      [ nan, ...,  nan]]]]])
Coordinates:
* lat        (lat) float64 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 ...
* lon        (lon) float64 260.0 261.0 262.0 263.0 264.0 265.0 266.0 267.0 ...
* time       (time) datetime64[ns] 1983-01-15T12:00:00
* latitude   (latitude) float32 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 ...
* longitude  (longitude) float32 -100.0 -99.0 -98.0 -97.0 -96.0 -95.0 
...
Attributes:
associated_files:  baseURL: http://cmip-
pcmdi.llnl.gov/CMIP5/dataLocation...
cell_methods:      time: mean
history:           2014-07-03T07:58:56Z altered by CMOR: Treated 
scalar d...
long_name:         Eastward Near-Surface Wind
standard_name:     eastward_wind
units:             m s-1

您可以看到，因为ws的维度名称为lon和lat，而ws的维度名称为经度和纬度，因此它创建的是5维的DataArray并且未正确拾取蒙版.

You can see because ws has dimension names lon and lat where as Hs has dimension names longitude and latitude it is creating a 5 dimension DataArray and not picking the mask up correctly.

无论尺寸名称不同还是DataArrays尺寸不同，我都可以选择蒙版吗?

Any way I can pick the mask regardless if the dimensions names are different or if the DataArrays are different sizes?

我以前是用numpy.math(ma)这样做的:

I previously did this with numpy.math (ma) as:

hs = f.variables['hs'][:]
hs_masked = ma.masked_values(hs, -65.534004)
tmp = np.zeros((len(lat), len(lon))
# Create masked array
data_cs = ma.masked_values(tmp, 0)
# Read new file
tmp = f.variables['cusp'][:]
data_cs[:,:] = ma.masked_array(tmp, hs_masked.mask)

但是希望学习/使用xarray.

But hoping to learn/use xarray.

干杯， 雷

推荐答案

您将需要显式重命名尺寸名称以进行匹配，例如hs = hs.rename({'lat': 'latitude', 'longitude': 'longitude'}).您可能还需要使用最近邻居索引，如果坐标标签不完全匹配，例如hs.reindex_like(ws10, method='nearest', tolerance=0.01).

You will need to explicitly rename dimension names to match, e.g., hs = hs.rename({'lat': 'latitude', 'longitude': 'longitude'}). You might also need to reindex with nearest-neighbor indexing, if the coordinate labels don't match exactly, e.g., hs.reindex_like(ws10, method='nearest', tolerance=0.01).

或者，出于不太安全的考虑，您可以从第二个参数中删除坐标标签，而只是传入未标记的数组，例如ws10.where(hs.data > 0).但我不建议使用此选项，因为没有任何方法可以保证元数据的一致性.

Or, less safely, you could strip out the coordinate labels from the second argument, and just pass in a unlabeled array instead, e.g., ws10.where(hs.data > 0). But I don't recommend this option, because nothing guarantees the consistency of the metadata.

这篇关于python-xarray从一个DataArray复制掩码到另一个的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！