[R-sig-Geo] uniformly sample points on a border of a polygon

[Tom Philippi]

Note that the spsurvey package can also do IRS (simple independent random
sample) on points, polylines (e.g., stream networks), and area (polygons).
While the irslin() function only shows shapefile and not sp object, the
irs() function allows you to specify an sp object.

So, you can convert your polygon boundary to an sp spatialLines object
boundary, then:

design1 <- list("Stratum 1"=list(panel=c(Panel=1000), seltype="Equal")

draw <- irs(design=design1, type.frame='linear',

           src.frame='sp.object', sp.object='boundary')  # I'm pretty
sure boundary is a quoted name

This is roughly the same design object & syntax for a grts draw.


Tom 2


On Fri, Oct 14, 2016 at 2:12 PM, mgm mgm <mgm917391 at gmail.com> wrote:

> Another possible solution is to use the spsurvey package, and think of each
> edge as a level of stratification and use the grts function to design a
> survey of sample points on the edges. An advantage of the grts function is
> that it uses a local neighborhood variance estimator, which can produce
> smaller confidence limits compared to simple random sample variiance.
> Mike
>
> On Friday, October 14, 2016, Paolo Piras <paolo.piras at uniroma3.it> wrote:
>
> > Dear Adrian,
> >
> > even this solution is pretty cool.
> >
> > Again, thanks to all who suggested me how to do that.
> >
> > Best
> >
> > Paolo
> >
> >
> > ________________________________
> > Da: Adrian Baddeley <adrian.baddeley at curtin.edu.au <javascript:;>>
> > Inviato: venerdì 14 ottobre 2016 02.14
> > A: Paolo Piras; Rolf Turner
> > Cc: r-sig-geo; Ege Rubak
> > Oggetto: Re: [FORGED] [R-sig-Geo] uniformly sample points on a border of
> a
> > polygon
> >
> >
> > You can use the spatstat function 'edges' to extract the edges of a
> > polygonal window.
> >
> >
> > Example:
> >
> >             W <- letterR    #polygonal window
> >
> >             E <- edges(W)
> >
> >             X <- runifpointOnLines(20, E)
> >
> >             plot(E)
> >
> >             plot(X, add=TRUE)
> >
> >
> >
> > Prof Adrian Baddeley DSc FAA
> >
> > Department of Mathematics and Statistics
> >
> > Curtin University, Perth, Western Australia
> >
> >
> > ________________________________
> > From: Paolo Piras <paolo.piras at uniroma3.it <javascript:;>>
> > Sent: Friday, 14 October 2016 5:22 AM
> > To: Rolf Turner
> > Cc: r-sig-geo; Adrian Baddeley; Ege Rubak
> > Subject: Re: [FORGED] [R-sig-Geo] uniformly sample points on a border of
> a
> > polygon
> >
> >
> > Thanks a lot Rolf!
> >
> > This is virtually exactly what I need;
> >
> > I'm very grateful for that
> >
> > All the best
> >
> > Paolo
> >
> >
> > ________________________________
> > Da: Rolf Turner <r.turner at auckland.ac.nz <javascript:;>>
> > Inviato: giovedì 13 ottobre 2016 23.12
> > A: Paolo Piras
> > Cc: r-sig-geo; Adrian.Baddeley at curtin.edu.au <javascript:;>; Ege Rubak
> > Oggetto: Re: [FORGED] [R-sig-Geo] uniformly sample points on a border of
> a
> > polygon
> >
> > On 14/10/16 07:03, Paolo Piras wrote:
> > > HI folks,
> > >
> > > I write for a (hopefully) relatively simple question:
> > >
> > > I would need to uniformly sample 1000 or more points **along the
> > border** of a polygon (not within the area enclosed) that is identified
> by
> > ordered but not equally spaced points; which is the fastest way?
> > >
> > > In a first moment I thought to sample between any pair of consecutive
> > points but, given that starting points are not uniformly distributed, the
> > final result would be very far from a uniform distribution.
> > >
> > > here my polygon:
> > >
> > >
> > >   mypol<-round(matrix(c(-13.8447497369687, -3.51439434200449,
> > 6.09494902836977, 6.83498916728338, 9.20403746769121, 15.3061452155498,
> > 18.4050681631565, 15.334153355932, 9.21809033073377, 6.90467983448734,
> > 6.17942233200763, -3.4864867866601, -13.8299219386242, -17.5237987124776,
> > -17.2262670680261, -17.5217563171495, -2.29667185082115,
> -7.72275721405543,
> > -9.77084968112857, -8.81725304021858, -8.32894043391822,
> -4.76080777897439,
> > -0.0600572363382094, 4.62779963258511, 8.20771806467615,
> 8.70484104396818,
> > 9.68531129857718, 7.67574865642846, 2.46081860449754, 1.31152149442131,
> > 0.0845735294613392, -1.11988475144136),ncol=2),digits=2)
> > >   plot(mypol,asp=1,cex=0)
> > >   text(mypol[,1],mypol[,2],c(1:nrow(mypol)))
> > > Thanks in advance for any hints
> >
> >
> > This can be done reasonably easily using the spatstat package, for some
> > value of the word "reasonably".  Here's how:
> >
> > require(spatstat)
> > W <- owin(poly=mypol)
> > m <- cbind(mypol[-nrow(mypol),],mypol[-1,])
> > m <- rbind(m,c(mypol[nrow(mypol),],mypol[1,]))
> > m <- as.data.frame(m)
> > names(m) <- c("x0","y0","x1","y1")
> > L <- with(m,psp(x0,y0,x1,y1,window=boundingbox(W)))
> > set.seed(42)
> > #X <- runifpointOnLines(1000,L)
> > X <- runifpointOnLines(100,L)
> > plot(W,main="Piras's Polygon")
> > plot(X,add=TRUE)
> >
> > Note that I have just generated 100 uniform points, r.t. 1000, so that
> > the resulting plot is a little less cluttered.
> >
> > There may be a sexier way of accomplishing your desideratum; I have
> > cc-ed this email to my co-authors Adrian and Ege who may come up with
> > better ideas.
> >
> > cheers,
> >
> > Rolf Turner
> >
> > --
> > Technical Editor ANZJS
> > Department of Statistics
> > University of Auckland
> > Phone: +64-9-373-7599 ext. 88276
> >
> >         [[alternative HTML version deleted]]
> >
> >
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

	[[alternative HTML version deleted]]