[GRASS-user] Interpolating rainfall data across area from points

[Rich Shepard]

On Thu, 11 Oct 2018, Thomas Adams wrote:

> Few (if any) of the spatial interpolation algorithms require "evenly
> spaced" data. Ultimately, you want spatial estimates that produce the
> least bias and where the errors are gaussian distributed, which can only
> be determined using some process of bootstrapping where stations are
> removed from the spatial interpolation process and are then used to
> evaluate error at those locations; and then, repeated, with a new set of
> stations. This should be done with several spatial interpolation
> algorithms that are generally known to produce the best estimates.

Tom,

   Hadn't thought of bootstrapping since I'm only now digging into geostats.

> If you are trying to produce estimates over an entire county, you really
> need stations well outside the county boundary to avoid having problems
> near the county periphry (interior of the county boundary).

   Good point. Many of the reporting stations are close to the border since
that's where the larger cities are located. Changing the analytical boundary
to put more stations outside it makes good sense. I can use a subwatershed
(the small bounded area in the attached map), or another defined area. Your
thoughts on how to define that area will help.

> GRASS and R play really well together through spgrass7, so that open the
> use of many R based geostatistical packages.

   I've used rgrass7 and was not aware of spgrass7. I'll install that
package.

Thanks,

Rich

-------------- next part --------------
A non-text attachment was scrubbed...
Name: map-wx-stations.jpg
Type: image/jpeg
Size: 116445 bytes
Desc: 
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20181011/753176fb/attachment-0001.jpg>