I want to apply mask on 2D numpy array. But it does not work correctly. Suppose I have
val(lat, lon) ---> my 2D array (20, 30)
Mask_lat = np.ma.masked_array(lat, mask=latmask) ---> masked lat (5,)
Mask_lon = np.ma.masked_array(lon, mask =lonmask) ---> masked lon (8,)
Maks_val = np.ma.masked_array(val, mask=mask_lat_lon) ---> ?
I do not know how can I pass a correct mask_lat_lon to have masked val (5,8). I would appreciate if one guides me.
Thank you in advance.
If I understand your question correctly, you have two 1D arrays that represent y and x (lat and long) positions in a 2D array. You want to mask a region based on the x/y position in the 2D array.
The key part to understand is that mask for a 2D array is also 2D.
For example, let's mask a single element of a 2D array:
import numpy as np
z = np.arange(20).reshape(5, 4)
mask = np.zeros(z.shape, dtype=bool)
mask[3, 2] = True
print z
print np.ma.masked_array(z, mask)
This yields:
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]
[12 13 14 15]
[16 17 18 19]]
[[ 0 1 2 3]
[ 4 5 6 7]
[ 8 9 10 11]
[12 13 -- 15]
[16 17 18 19]]
In your case, you have two 1D x and y arrays that you need to create a 2D mask from. For example:
import numpy as np
x = np.linspace(-85, -78, 4)
y = np.linspace(32, 37, 5)
z = np.arange(20).reshape(5, 4)
xmask = (x > -82.6) & (x < -80)
ymask = (y > 33) & (y < 35.6)
print xmask
print ymask
We'd then need to combine them into a single 2D mask using broadcasting:
mask = xmask[np.newaxis, :] & ymask[:, np.newaxis]
Slicing with newaxis (or None, they're the same object) adds a new axis at that position, turning the 1D array into a 2D array. It you have seen this before, it's useful to take a quick look at what xmask[np.newaxis, :] and ymask[:, np.newaxis] look like:
In [14]: xmask
Out[14]: array([False, False, True, False], dtype=bool)
In [15]: ymask
Out[15]: array([False, True, True, False, False], dtype=bool)
In [16]: xmask[np.newaxis, :]
Out[16]: array([[False, False, True, False]], dtype=bool)
In [17]: ymask[:, np.newaxis]
Out[17]:
array([[False],
[ True],
[ True],
[False],
[False]], dtype=bool)
mask will then be (keep in mind that True elements are masked):
In [18]: xmask[np.newaxis, :] & ymask[:, np.newaxis]
Out[18]:
array([[False, False, False, False],
[False, False, True, False],
[False, False, True, False],
[False, False, False, False],
[False, False, False, False]], dtype=bool)
Finally, we can create a 2D masked array from z based on this mask:
arr = np.masked_array(z, mask)
Which gives us our final result:
[[ 0 1 2 3]
[ 4 5 -- 7]
[ 8 9 -- 11]
[12 13 14 15]
[16 17 18 19]]
mask as shown above, then clipped_z = z[mask]. This creates a 1D array that can then be converted to 2D with clipped_z.reshape(). Appropriate dimensions for reshaping can be obtained with np.sum(xmask) and np.sum(ymask), which count the number of True.
lonmask&latmaskbut I don't understand precisely what ismasked lat (5,)