[GRASS-user] r.neighbor circular mask VS r.buffer on single pixel

[Enrico Gallo]

2012/9/6 Enrico Gallo <enrico.gallo at gmail.com>:

> To begin, I used r.neighbor with a circular mask , size 41 pixel.
> I repeated a similar analysis (adding another filter) parsing r.univar
> on each 1km buffer around each pixel (using r.mask)
>
> Just for the sake of precision, I also tested my routine not
> considering the added filter and I notice a few cases where the two
> results slightly differ, probably due a difference between r.neighbor
> circular mask and r.buffer output used as mask

Solved

The problem was not in r.neighbor nor in r.buffer, but in a
preprocessing step I made (and I omitted in the previous message).
In the first case (r.neighbor) I limited computation only on pixels of
the channel network extracted by r.stream.extract
In the second one (r.buffer + r.mask) for each pixel I used downstream
channel network extracted by r.drain (on the same accumulation map
used as input of r.stream.extract, but inverted)

The two networks "slightly differ" , and local statistics too.
Aligning the two network, all was fine.


Enrico