使用 Python 3 的 NetCDF 时间序列切片

Question

我正在尝试从 NetCDF 文件中绘制一周的时间序列数据，但遇到了一些问题。

我正在使用以下软件包：

import netCDF4
from matplotlib import pyplot as plt
import numpy as np
import xarray as xr
import dask

首先我导入两个 .nc 文件：

ds1 = xr.open_dataset('ERA5_forecast_100V_247_2008.nc')
ds2 = xr.open_dataset('ERA5_analysis_100V_247_2008.nc')

然后我使用 xarray 选择时间和网格位置：

dsloc1 = ds1.sel(time='2008-02-10',longitude=2.2,latitude=48.7,method='nearest')
dsloc2 = ds2.sel(time='2008-02-10',longitude=2.2,latitude=48.7,method='nearest')

然后我绘制两个时间序列：

dsloc1['v100'].plot.line('b-',figsize=(15,10))
dsloc2['v100'].plot.line('y-')

这会产生我所期望的：

但是，当我尝试选择一个日期范围时，我得到了一些错误...

dsloc1 = ds1.sel(time=slice('2008-03-01','2008-03-07'),longitude=2.2,latitude=48.7,method='nearest')
dsloc2 = ds2.sel(time=slice('2008-03-01','2008-03-07'),longitude=2.2,latitude=48.7,method='nearest')

我确定这可能是语法问题，但我花了比我想要的时间更长的时间来尝试解决它。​​任何建议都非常感谢！

[编辑] 这是回溯：

---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
<ipython-input-308-c3385fa732ab> in <module>()
      1 # select time and grid location (Feb 10th, 2008, near Paris)
----> 2 dsloc1 = ds1.sel(time=slice('2008-03-01','2008-03-07'),longitude=2.2,latitude=48.7,method='nearest')
      3 dsloc2 = ds2.sel(time=slice('2008-03-01','2008-03-07'),longitude=2.2,latitude=48.7,method='nearest')

/usr/local/lib/python3.6/site-packages/xarray/core/dataset.py in sel(self, indexers, method, tolerance, drop, **indexers_kwargs)
   1507         indexers = either_dict_or_kwargs(indexers, indexers_kwargs, 'sel')
   1508         pos_indexers, new_indexes = remap_label_indexers(
-> 1509             self, indexers=indexers, method=method, tolerance=tolerance)
   1510         result = self.isel(indexers=pos_indexers, drop=drop)
   1511         return result._replace_indexes(new_indexes)

/usr/local/lib/python3.6/site-packages/xarray/core/coordinates.py in remap_label_indexers(obj, indexers, method, tolerance, **indexers_kwargs)
    353 
    354     pos_indexers, new_indexes = indexing.remap_label_indexers(
--> 355         obj, v_indexers, method=method, tolerance=tolerance
    356     )
    357     # attach indexer's coordinate to pos_indexers

/usr/local/lib/python3.6/site-packages/xarray/core/indexing.py in remap_label_indexers(data_obj, indexers, method, tolerance)
    248         else:
    249             idxr, new_idx = convert_label_indexer(index, label,
--> 250                                                   dim, method, tolerance)
    251             pos_indexers[dim] = idxr
    252             if new_idx is not None:

/usr/local/lib/python3.6/site-packages/xarray/core/indexing.py in convert_label_indexer(index, label, index_name, method, tolerance)
    132         if method is not None or tolerance is not None:
    133             raise NotImplementedError(
--> 134                 'cannot use ``method`` argument if any indexers are '
    135                 'slice objects')
    136         indexer = index.slice_indexer(_sanitize_slice_element(label.start),

NotImplementedError: cannot use ``method`` argument if any indexers are slice objects

Answer 1

似乎根本不支持使用时间片结合method='nearest' 的sel：

如果任何索引器是切片对象，则不能使用 method 参数

这在某种程度上是有道理的，因为为切片选择 nearest 似乎有点奇怪。

您可以通过分两步执行sel 来解决此问题，即首先选择时间片，然后从该时间片中选择一个位置（或相反）。我不确定这是否是最好的解决方案，但至少它有效。

带有一些 ERA5 数据的快速示例：

import xarray as xr

ds1 = xr.open_dataset('20160502_cabauw_model_fc.nc')

# Works:
dsloc1 = ds1.sel(time='2016-05-02 10:00', longitude=4.9, latitude=51.2, method='nearest')

# Doesn't work:
#dsloc2 = ds1.sel(time=slice('2016-05-02 10:00', '2016-05-02 12:00'), longitude=4.9, latitude=51.2, method='nearest')

# Works:
tmp    = ds1.sel(time=slice('2016-05-02 10:00', '2016-05-02 12:00'))
dsloc2 = tmp.sel(longitude=4.9, latitude=51.2, method='nearest')

结果如下：

In [23]: dsloc2
Out[23]: 
<xarray.Dataset>
Dimensions:    (level: 137, time: 3)
Coordinates:
    longitude  float32 4.8
    latitude   float32 51.3
  * level      (level) int32 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 ...
  * time       (time) datetime64[ns] 2016-05-02T10:00:00 2016-05-02T11:00:00 ...
Data variables:
    z          (time, level) float32 ...