Basemap interp nearest neighbor

Juan: I agree that this would be desirable behavior. Unfortunately, it's not obvious to me how to do it. I'll think about it and get back to you. (cc'ing matplotlib-users list).

-Jeff

···

On 3/6/11 8:58 PM, Juan A. Saenz wrote:

Hi,

I use Basemap and netCDF4-python on a regular basis, and find them very useful tools. Thank you for developing them!

I noticed that when using basemap.interp for nearest neighbor (order=0) the interpolation is not masked, and nearest neighbor masked values will be used in the interpolation. I was wondering if you could suggest a way to do nearest neighbor interpolation where masked are supported, i.e. nearest neighbor values that are not masked.

Thanks for your help,
Juan

Juan: On second thought, I'm not sure this is desirable behavior. I would guess that most of the time, if a nearest neighbor is masked, the user would expect the interpolation routine to return a masked value. I would be interested to hear what others think.

-Jeff

···

On 3/7/11 5:50 AM, Jeff Whitaker wrote:

On 3/6/11 8:58 PM, Juan A. Saenz wrote:

Hi,

I use Basemap and netCDF4-python on a regular basis, and find them
very useful tools. Thank you for developing them!

I noticed that when using basemap.interp for nearest neighbor
(order=0) the interpolation is not masked, and nearest neighbor masked
values will be used in the interpolation. I was wondering if you could
suggest a way to do nearest neighbor interpolation where masked are
supported, i.e. nearest neighbor values that are not masked.

Thanks for your help,
Juan

Juan: I agree that this would be desirable behavior. Unfortunately,
it's not obvious to me how to do it. I'll think about it and get back
to you. (cc'ing matplotlib-users list).

-Jeff

On a similar note, are there any alternatives available to nearest neighbor? For example, kriging? I remember seeing a geostats library in python (hpgl i think), but I found the API rather impractical and difficult to use.

Thanks,
Aman

···

On Mon, Mar 7, 2011 at 8:23 AM, Jeff Whitaker <jswhit@…146…> wrote:

On 3/7/11 5:50 AM, Jeff Whitaker wrote:

On 3/6/11 8:58 PM, Juan A. Saenz wrote:

Hi,

I use Basemap and netCDF4-python on a regular basis, and find them

very useful tools. Thank you for developing them!

I noticed that when using basemap.interp for nearest neighbor

(order=0) the interpolation is not masked, and nearest neighbor masked

values will be used in the interpolation. I was wondering if you could

suggest a way to do nearest neighbor interpolation where masked are

supported, i.e. nearest neighbor values that are not masked.

Thanks for your help,

Juan

Juan: I agree that this would be desirable behavior. Unfortunately,

it’s not obvious to me how to do it. I’ll think about it and get back

to you. (cc’ing matplotlib-users list).

-Jeff

Juan: On second thought, I’m not sure this is desirable behavior. I

would guess that most of the time, if a nearest neighbor is masked, the

user would expect the interpolation routine to return a masked value. I

would be interested to hear what others think.

-Jeff


What You Don’t Know About Data Connectivity CAN Hurt You

This paper provides an overview of data connectivity, details

its effect on application quality, and explores various alternative

solutions. http://p.sf.net/sfu/progress-d2d


Matplotlib-users mailing list

Matplotlib-users@lists.sourceforge.net

https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Aman: The basemap interp function is just a convenience function
for simple regridding (linear, cubic and nearest neighbor).
scipy.interpolate
()
provides many more options.
-Jeff

···

http://docs.scipy.org/doc/scipy/reference/interpolate.html

    On Mon, Mar 7, 2011 at 8:23 AM, Jeff

Whitaker <jswhit@…146…>
wrote:

On 3/7/11 5:50 AM, Jeff Whitaker wrote:

        > On 3/6/11 8:58 PM, Juan A. Saenz wrote:

        >> Hi,

        >>

        >> I use Basemap and netCDF4-python on a regular

basis, and find them

        >> very useful tools. Thank you for developing them!

        >>

        >> I noticed that when using basemap.interp for

nearest neighbor

        >> (order=0) the interpolation is not masked, and

nearest neighbor masked

        >> values will be used in the interpolation. I was

wondering if you could

        >> suggest a way to do nearest neighbor interpolation

where masked are

        >> supported, i.e. nearest neighbor values that are

not masked.

        >>

        >> Thanks for your help,

        >> Juan

        >>

        > Juan:  I agree that this would be desirable behavior.

Unfortunately,

        > it's not obvious to me how to do it.  I'll think about

it and get back

        > to you.  (cc'ing matplotlib-users list).

        >

        > -Jeff

        >
      Juan: On second thought, I'm not sure this is desirable

behavior. I

      would guess that most of the time, if a nearest neighbor is

masked, the

      user would expect the interpolation routine to return a masked

value. I

      would be interested to hear what others think.


          -Jeff

          What You Don't Know About Data Connectivity CAN Hurt You

          This paper provides an overview of data connectivity,

details

          its effect on application quality, and explores various

alternative

          solutions. [http://p.sf.net/sfu/progress-d2d](http://p.sf.net/sfu/progress-d2d)

          _______________________________________________

          Matplotlib-users mailing list

          Matplotlib-users@lists.sourceforge.net

          [https://lists.sourceforge.net/lists/listinfo/matplotlib-users](https://lists.sourceforge.net/lists/listinfo/matplotlib-users)

Jeff, thanks for your reply.

One situation where one might require masked nearest neighbor interpolation is when, on a given fixed grid, interpolating velocities on cell corners (B-grid) to faces (C-grid). Cells will be defined as either land or ocean cells, masked or un-masked respectively. The masked or un-masked character of cells does not change. Interpolating velocities from corners adjacent to masked cells, to cell centers on un-masked cells will require the behavior in question. Imagine two adjacent cells, one masked and the other not. The velocities on the cell corners that lie on the coast (adjacent to a masked and an un-masked cell) are masked. A desireable behavior of the interpolator would be to produce an un-masked cell centered velocity on the un-masked cell that uses values from adjacent un-masked velocity values.

Thanks,
Juan

···

On 8/03/11 12:23 AM, Jeff Whitaker wrote:

On 3/7/11 5:50 AM, Jeff Whitaker wrote:

On 3/6/11 8:58 PM, Juan A. Saenz wrote:

Hi,

I use Basemap and netCDF4-python on a regular basis, and find them
very useful tools. Thank you for developing them!

I noticed that when using basemap.interp for nearest neighbor
(order=0) the interpolation is not masked, and nearest neighbor masked
values will be used in the interpolation. I was wondering if you could
suggest a way to do nearest neighbor interpolation where masked are
supported, i.e. nearest neighbor values that are not masked.

Thanks for your help,
Juan

Juan: I agree that this would be desirable behavior. Unfortunately,
it's not obvious to me how to do it. I'll think about it and get back
to you. (cc'ing matplotlib-users list).

-Jeff

Juan: On second thought, I'm not sure this is desirable behavior. I would guess that most of the time, if a nearest neighbor is masked, the user would expect the interpolation routine to return a masked value. I would be interested to hear what others think.

-Jeff

--
Juan A. Saenz
Postdoctoral Fellow

Geophysical Fluid Dynamics
Research School of Earth Sciences
Australian National University

Building 61
Mills Road
Canberra, ACT 0200
AUSTRALIA
Office: +61 2 6125 9968
Admin: +61 2 6125 5502
Fax: +61 2 6257 2737

Jeff, thanks for your reply.

One situation where one might require masked nearest neighbor interpolation is when, on a given fixed grid, interpolating velocities on cell corners (B-grid) to faces (C-grid). Cells will be defined as either land or ocean cells, masked or un-masked respectively. The masked or un-masked character of cells does not change. Interpolating velocities from corners adjacent to masked cells, to cell centers on un-masked cells will require the behavior in question. Imagine two adjacent cells, one masked and the other not. The velocities on the cell corners that lie on the coast (adjacent to a masked and an un-masked cell) are masked. A desireable behavior of the interpolator would be to produce an un-masked cell centered velocity on the un-masked cell that uses values from adjacent un-masked velocity values.

Thanks,
Juan

Juan: I can see why you want it in this case, but I think in general users would expect a masked value if the nearest neighbor is masked. In addition, I don't see how to implement it easily. How far are you willing to go to find a nearest neighbor that is not masked?

In short, for your use case you'll have to implement your own custom solution. Of course, if you can show me a simple modification to the Basemap interp function that does what you want, and can be enabled with a kwarg, I'll reconsider.

-Jeff

···

On 3/7/11 2:25 PM, Juan A. Saenz wrote:

On 8/03/11 12:23 AM, Jeff Whitaker wrote:

On 3/7/11 5:50 AM, Jeff Whitaker wrote:

On 3/6/11 8:58 PM, Juan A. Saenz wrote:

Hi,

I use Basemap and netCDF4-python on a regular basis, and find them
very useful tools. Thank you for developing them!

I noticed that when using basemap.interp for nearest neighbor
(order=0) the interpolation is not masked, and nearest neighbor masked
values will be used in the interpolation. I was wondering if you could
suggest a way to do nearest neighbor interpolation where masked are
supported, i.e. nearest neighbor values that are not masked.

Thanks for your help,
Juan

Juan: I agree that this would be desirable behavior. Unfortunately,
it's not obvious to me how to do it. I'll think about it and get back
to you. (cc'ing matplotlib-users list).

-Jeff

Juan: On second thought, I'm not sure this is desirable behavior. I would guess that most of the time, if a nearest neighbor is masked, the user would expect the interpolation routine to return a masked value. I would be interested to hear what others think.

-Jeff

--
Jeffrey S. Whitaker Phone : (303)497-6313
Meteorologist FAX : (303)497-6449
NOAA/OAR/PSD R/PSD1 Email : Jeffrey.S.Whitaker@...259...
325 Broadway Office : Skaggs Research Cntr 1D-113
Boulder, CO, USA 80303-3328 Web : http://tinyurl.com/5telg

If I think of something, I'll let you know.

In the meantime, I'd like to point out the following:
Nearest neighbor returns a masked interpolation point if the nearest neighbor is masked (this is just what you've already told me). If there are two equidistant neighbors, it returns the one on the "bottom". I, naively, expected it to return the unmasked one, but I suppose that this is just natural. It might be worth pointing this out in the documentation.

Thanks again,
Juan

···

On 8/03/11 10:52 AM, Jeff Whitaker wrote:

On 3/7/11 2:25 PM, Juan A. Saenz wrote:

Jeff, thanks for your reply.

One situation where one might require masked nearest neighbor interpolation is when, on a given fixed grid, interpolating velocities on cell corners (B-grid) to faces (C-grid). Cells will be defined as either land or ocean cells, masked or un-masked respectively. The masked or un-masked character of cells does not change. Interpolating velocities from corners adjacent to masked cells, to cell centers on un-masked cells will require the behavior in question. Imagine two adjacent cells, one masked and the other not. The velocities on the cell corners that lie on the coast (adjacent to a masked and an un-masked cell) are masked. A desireable behavior of the interpolator would be to produce an un-masked cell centered velocity on the un-masked cell that uses values from adjacent un-masked velocity values.

Thanks,
Juan

Juan: I can see why you want it in this case, but I think in general users would expect a masked value if the nearest neighbor is masked. In addition, I don't see how to implement it easily. How far are you willing to go to find a nearest neighbor that is not masked?

In short, for your use case you'll have to implement your own custom solution. Of course, if you can show me a simple modification to the Basemap interp function that does what you want, and can be enabled with a kwarg, I'll reconsider.

-Jeff

On 8/03/11 12:23 AM, Jeff Whitaker wrote:

On 3/7/11 5:50 AM, Jeff Whitaker wrote:

On 3/6/11 8:58 PM, Juan A. Saenz wrote:

Hi,

I use Basemap and netCDF4-python on a regular basis, and find them
very useful tools. Thank you for developing them!

I noticed that when using basemap.interp for nearest neighbor
(order=0) the interpolation is not masked, and nearest neighbor masked
values will be used in the interpolation. I was wondering if you could
suggest a way to do nearest neighbor interpolation where masked are
supported, i.e. nearest neighbor values that are not masked.

Thanks for your help,
Juan

Juan: I agree that this would be desirable behavior. Unfortunately,
it's not obvious to me how to do it. I'll think about it and get back
to you. (cc'ing matplotlib-users list).

-Jeff

Juan: On second thought, I'm not sure this is desirable behavior. I would guess that most of the time, if a nearest neighbor is masked, the user would expect the interpolation routine to return a masked value. I would be interested to hear what others think.

-Jeff