[AniMov] overlap

[Clément Calenge]

Hi all,


> The result is a non-symmetric distance matrix, which eventually can be 
> aggregated to have, say, average overlaps between sexes, seasons, etc..
>
> In practice, all that is done with a slightly modified version of Clément's 
> kerneloverlap. the modification consists in commenting put the first rows 
> that call kernelUD, and making function kerneloverlap work on already 
> calculated UDs.
>
> This raises another point, i.e. a suggestion to Clément: since often the 40x40 
> cells approach gives undesirable results, we prefer to prepare as a 
> preliminary step, a raster (of course using exclusively ESRI software, duh!) 
> with cells of suitable size, that encompassess an extent slightly wider than 
> the study area, an then calculate once all the UDs.
>   

Note that controlling the grid can be done with kerneloverlap: this 
functions accepts all the arguments of kernelUD. For example, using the 
puechabon dataset:

data(puechabon)

## With 200x200 grid
kerneloverlap(puechabon$locs[,c("X","Y")], puechabon$locs$Name,
                        grid=200, meth="VI", conditional=TRUE)

## With a map available in R
map <- getkasc(puechabon$kasc,1)
kerneloverlap(puechabon$locs[,c("X","Y")], puechabon$locs$Name,
                        grid=map, meth="VI", conditional=TRUE)


> This 'single point of homerange truth' [2] guarantees a consistency between 
> subsequant calculations, i.e. 'getvoumes + polygon extraction', compana, 
> kerneloverlapd etc.
>
> This also saves lots of time, since UD calculation, moreover with a 'fine' 
> grid, takes some time: UD processing can be done on a fast number crunching 
> machine [3], and the result can be saved (as part of a sahrlocs structure) 
> and retrieved on a less performant host to do other postprocessing.
>   

Yes, you are right: kernel UDs are often computed as a basis for further 
analysis, including kernel overlap, but not only... It would be more 
sensible to estimate UD once, and then using them for other analyses. I 
planned to upload a new version of adehabitat to CRAN before the end of 
the week, and I will include, on the help page of kerneloverlap, a new 
function accepting a "khrud" object to perform this computation.
Many thanks for the suggestion,
Regards,

Clément

-- 
Clément CALENGE
Office national de la chasse et de la faune sauvage
Saint Benoist - 78610 Auffargis
tel. (33) 01.30.46.54.14