Helper Function to determine resolution of coordinates

[mpu-creare]

Description When working with lat/lon or other projected coordinates, it's hard to know the real nominal resolution of your coordinate system.

Describe the solution you'd like I want a helper function on coordinates that give me the resolution.

Sample Code

>>> coords = podpac.Coordinates([...], dims=['lat', 'lon']
>>> coords.horizontal_resolution(units='m', type='nominal')
{"lat": 10,
"lon": 20, 
"units": "m"
}

[mpu-creare]

This code might be useful: https://github.com/creare-com/pydem/blob/master/pydem/utils.py#L145

[mpu-creare]

This might also be useful: https://pyproj4.github.io/pyproj/stable/api/crs/crs.html

[CFoye-Creare]

Left to-do

• [x] Finish Unit Tests
• [x] Check for units

[CFoye-Creare]

• [x] cKDTree with boxsize wraparound implementation for stacked NearestNeighbor
• [x] UTM projection handling

[mpu-creare]

For coords.horizontal_resolution(units='m', type='nominal'), let's talk about the different types

types =

• "nominal" <-- this is wrong but cheap to calculate. Give a 'nominal' resolution over the entire domain
• "summary" <-- This is still approximate, gives the mean and standard deviation to each point
• "full" <-- min distance between every point (dx, dy, basically)
• "full_exact" <-- uses brute-force for the stacked coordinates

This is a proposal, feel free to propose something better.

[CFoye-Creare]

Implemented 3 of the 4 types:

• "nominal" <-- this is wrong but cheap to calculate. Give a 'nominal' resolution over the entire domain
• "summary" <--- Returns average and std.dev for each point.
• "full" <--- returns the exact distance between every point, or every dx,dy if coordinates are unstacked

[CFoye-Creare]

9851b1cb5ce4b32bc00e96e500ffc8c78efd7376

[CFoye-Creare]

e05a662688b5d600aeb29784d43b5acb8c466cbe

[CFoye-Creare]

Wrote some code to address the problem of lat/lon evaluating to different distances based upon the other dimension's value. However, I realized once passing in non-square unstacked coordinates that I failed to account for the difference resolution changing for each differing value of the other dimension.

Let our unstacked coordinate system be a $M$ by $N$ matrix, where $M$ is the number of rows (in grid squares), and $N$ is the number of columns. See Blender visualization:

Previously, we were returning two arrays of size $M$ and size $N$. This meant we were returning either a single row/column or a selection of each column. For example: Another example:

Instead, we want to return this for latitude:

So, for latitude we want to return a $N$ by $M$ matrix.

For longitude: So, for longitude we want to return a $M$ by $N$ matrix.

For the "summary" type resolution, we can simply average/compute std.dev of all of these values.

[CFoye-Creare]

The $M+1$ or $N+1$ implementation is not possible due to the fact we don't store the values in the +1 dimension in our coords.

[mpu-creare]

closed by #505