Spatial indepedence and covariate/buffer intersection (edit)

[JDJC]

Hi folks,

[*]Edit. Maybe the best way to ask this is as such: Are there statistical constraints in terms of building a use/non-use or presence/absence model based upon a moving window of explanatory covariates? In other words, can two units consider portions of the same "patch" if the specific surrounding configuration of the landscape differs (or, can explanatory buffers around detection stations have some overlap/spatial intersection and be considered truly independent?)

Cheers,

John

[bacollier]

Hi folks,
[*]Edit. Maybe the best way to ask this is as such: Are there statistical constraints in terms of building a use/non-use or presence/absence model based upon a moving window of explanatory covariates? In other words, can two units consider portions of the same "patch" if the specific surrounding configuration of the landscape differs (or, can explanatory buffers around detection stations have some overlap/spatial intersection and be considered truly independent?)

Cheers,
John

John,
I think it depends upon the scale of the window. I applied an occupancy model that used a covariate (woodland composition) that was based on a moving window used to estimate the proportion of woodland within a 400m circle to the pixel of interest (kind of a landscape contiguity metric), which we then averaged at the patch scale to provide a patch level values for 60K odd patches across 35 counties. In that case, I think its fine. But if you are thinking about a moving window of 400m and you are predicting to a 100m circle, then I might have more concerns.

bret