I have a 1 dimensional array of CO2 values where I need to repeat each value over lon/lat dimensions [360, 720] creating an array of dims [365, 360, 720] - i.e. 1 year of daily CO2 values across longitude and latitude.

I would want to take an array:

a = np.array([343.79258065, 343.79096774, 343.78935484])

And tile the first value across an array of dims [360, 720], then tile the second value across the same dims [360, 720], and do that for n values in the array (365 times in my case)

A small dim example of this would be (but please note the dims I want below):

array([[343.79258065, 343.79258065, 343.79258065, ...],
[343.79096774, 343.79096774, 343.79096774, ...],
[343.78935484, 343.78935484, 343.78935484, ...]])

The output dimensions

So effectively each value in array a would be repeated (tiled?) over an array of dims [360, 720] for 365 layers resulting in a 3D array of dims [365, 360, 720].

It would be great to solve this with broadcast_to() if it's possible. Some links that don't quite do what I want: Repeating values n times and Repeat across multiple dimensions

CodePudding user response：

I hope I understand your question correctly! I would do this by first reshaping the array into (N,1,1), then expanding each of those 1x1 sub-arrays into the repeated arrays you wanted to be (360, 720) using numpy's np.repeat function. Here's how I did it:

a = np.array([343.79258065, 343.79096774, 343.78935484]).reshape(-1,1,1)
a = np.repeat(a, 360, axis=1)
a = np.repeat(a, 720, axis=2)
print(a.shape)
>> (3, 360, 720)

If this isn't what you had in mind, please let me know! Each of the N 360x720 arrays contains only repeated instances of the ith value of the initial array (a[i]), by the way.