Dear Stuart,
I am hampered from a very careful assessment of your problem by being partially colorblind, but I have a fair
amount of experience with the Barnes analysis scheme. It is capable of producing grid point values that are
beyond the bounds of the data. Moreover, it suffers--as many other interpolation schemes do--from irregular
data density. There are profound edge effects at the edges of data, but many other problems happen even
internal to the data space when the observations are dense in some areas and sparse in others. Your sample
is a parade example of that! The clustered observations count more just because there are more of them,
until you get to a grid point very close to an isolated observation. Then, suddenly that observation becomes
much more important because the distance weight becomes very large. This causes strange things to happen
when you then go to the second, correction pass through the data.
I confess I have not used the IDV Barnes scheme much so there may be other
issues.
Dave Barber
On Mar 24, 2011, at 11:06 AM, Stuart Wier <wier@xxxxxxxxxx> wrote:
A test of the Barnes analysis in the IDV gives odd results.
The first image here shows the point observations as dots colored by value.
The second image adds in a grid made with the Barnes analysis in the IDV. Both
data use the same color scale and range.
Note the poor grid values in some places, such as the pale region of positive value in the right side, center, where
there are no positive values in the observations anywhere nearby.
Any suggestions about what is happening or how to get a better result will be
appreciated.
S K Wier
UNAVCO
<IDV-Barnes-test-data-pts.jpg>
<IDV-Barnes-test-data-grid.jpg>
_______________________________________________
idvusers mailing list
idvusers@xxxxxxxxxxxxxxxx
For list information, to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
