numpy.searchsorted in a multidimensional array

Question

I have got a three dimensional array with air pressure values in the form:

[[[1000 1010]
  [1005 990]]

[[950 960]
 [955 940]]

[[900 910]
[905 890]]]

The structure represents the pressure at different levels, so each element in the 2-d is ordered for each layer.

I would like to know at which level the pressure is 950 would be for each 2d element, getting a 2-d array with the index of the level for each element.

In a 1-D array like

a = [890, 940, 990]

I would use

a.searchsorted(950)

and the result would be 2, indicating that 950 would go at the 3rd position.

Is there a way to do it for all my array at once, without having to do it for each 2-d element?

I don't totally understand what you're after. If you take a single (m,n) array of pressure values corresponding to one level, are you saying that you want the rank of value v within the (m*n,) vector of sorted values for that level, e.g. np.sort(A[0,:,:].flat).searchsorted(v)? Then you just want to do the same thing for each level? — ali_m
– ali_m, Commented Jul 19, 2013 at 12:57
Well, I have multiple (m,n) arrays, one for each layer. I want to know between which levels a value would be, for each element in (m,n). i.e. For the element [1][1], the value 950 would be between the layer 0 and the layer 1. — Roger Veciana
– Roger Veciana, Commented Jul 19, 2013 at 13:23
I see, so it must be the case that for a given point A[:,i,j] the pressure values increase monotonically across levels (otherwise what you're asking for doesn't make sense) — ali_m
– ali_m, Commented Jul 19, 2013 at 13:26
Yes. Since they are atmospheric pressures, the values are accumulative for each point in the surface. — Roger Veciana
– Roger Veciana, Commented Jul 19, 2013 at 13:29
If you can use PyTorch, it supports N-D arrays: pytorch.org/docs/stable/generated/torch.searchsorted.html — Thomas Ahle
– Thomas Ahle, Commented Jul 20, 2022 at 18:28

w-m · Accepted Answer · 2013-07-19 13:33:24Z

6

You can apply the searchsorted function along an axis of your input array like this:

numpy.apply_along_axis(lambda a: a.searchsorted(950), axis = 1, arr = air_pr)

which should yield the intended result if I understand you correctly.

edited Jul 19, 2013 at 13:33

answered Jul 19, 2013 at 12:54

w-m

11.3k1 gold badge46 silver badges51 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Roger Veciana Over a year ago

Thank you! It gave me the solution, which is numpy.apply_along_axis(lambda a: a.searchsorted(950), axis = 0, arr = air_pressure) The array must be passed with arr =, and the axis is 0 (I still have to understand why). Also, the pressure values have to be ordered from lower to higher, but I think I will find a solution for that.

w-m Over a year ago

Ok. I had removed the 'arr =' to fit the command into one line, but of course Python will complain a non-keyword arg at that position.

Collectives™ on Stack Overflow

numpy.searchsorted in a multidimensional array

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related