In quassinja/mytestpkg: Demonstration of concepts

knitr::opts_chunk$set(echo = TRUE) #include=FALSE nicht in pdf output?

Point Pattern

harran=read.table("../data/Sites_HarranPlain.csv",
                  sep = ",",
                  header = TRUE) # when knitting: "../data/Sites_HarranPlain.csv"!!!!!
str(harran)

spatstat

library(sp)
coordinates(harran) <- ~X+Y
proj4string(harran) <- CRS("+init=epsg:4326")

harran <- spTransform(harran, CRSobj = CRS("+init=epsg:32637"))
str(harran) # for checking


library(spatstat)

str(harran@coords) # structure

harran_ppp <- ppp(x=harran@coords[,1],
                  y=harran@coords[,2],
                  window = owin(xrange = harran@bbox[1,],
                                yrange = c(min(harran@coords[,2]),
                                          min(harran@coords[,2])+52000)))


harran_ppp=unique.ppp(harran_ppp) # shows number of duplicated points and deletes them/ harran_ppp= has to be done to define harran_ppp new

str(harran_ppp)
plot(harran_ppp)

library(mapview)
mapview(harran)

Challenge: delete duplicated points

harran_ppp=unique.ppp(harran_ppp) # shows number of duplicated points and deletes them/ harran_ppp= has to be done to define harran_ppp new


# or:
#anyDuplicated(harran_ppp)
#harran <- unique(harran_ppp)
#harran_ppp <- harran_ppp[!duplicated(harran_ppp)]

plot(harran_ppp)

Nearest neighbour distance

harran_ppp_nn <- nndist(harran_ppp)
str(harran_ppp_nn) # shows distance within the structure(str)

hist(harran_ppp_nn)  # plots the nearest neighbour
#barplot(sort(harran_ppp_nn))

challenge: create a kernel density estimation

harran_kde <- density.ppp(harran_ppp,sigma = mean(harran_ppp_nn))# see: likelihood cross validation bandwidth selection for kernel density (help)
plot(harran_kde)

raster

library(raster)
dem <- raster("../data/dem.tif") # see above for problems when knitting

# or: library(rgdal)
#dem <- readGDAL("data/dem.tif")
plot(dem)

im_dem <- as.im(as.image.SpatialGridDataFrame(as(dem,"SpatialGridDataFrame"))) #creates image
plot(im_dem)

challenge: use rhohat and create a plot

#?rhohat # smoothing estimate: changes the raster
harran_rhohat <- rhohat(harran_ppp,im_dem,bw = 200)
              # <- rhohat(harran_ppp, im_dem, bw=200) /gives a more distinct picture

plot(harran_rhohat) #x=elevation y=relative intensity of points -> relation of elevation to pointdensity, bandwidth=default -> default=sigma in the structure of the object(str(harran_rhohat))

rho_dem <- predict(harran_rhohat)
plot(rho_dem)

diff_rho <- harran_kde-rho_dem
plot(diff_rho)

challenge: test poisson, create random points with rpoispp function that have the same intensity like our points

set.seed(123)
harran_rpoispp2 <- rpoispp(lambda = harran_ppp$n/area.owin(harran_ppp$window),
                          win=harran_ppp$window)
set.seed(123)
harran_rpoispp3 <- rpoispp(intensity(harran_ppp),win=Window(harran_ppp))
set.seed(123)
harran_rpoispp4 <- rpoispp(ex = harran_ppp)

plot(harran_ppp)
points(harran_rpoispp2,col="green")
points(harran_rpoispp3,col="blue")
points(harran_rpoispp4,col="red")

# first block is all the same, different ways to get the same result

Second order effects

harran_g <- Gest(harran_ppp)
str(harran_g)
plot(harran_g) # x=closest neighbours expected (blue), the rest shows higher than expected clusters y= distance

harran_ge <- envelope(harran_ppp,fun = "Gest") # calculates g function for random points
plot(harran_ge) # grey shadow_ monte Carlo Simulation

challenge: do F and K Function

#F-function:

harran_f <- Fest(harran_ppp)
plot(harran_f)

harran_fe <- envelope(harran_ppp,fun = "Fest") # calculates f function for random points
plot(harran_fe)

## red: expected, black deviates -> expect that the empty spaces are smaller than expected = clustered


#K-function

harran_k <- Kest(harran_ppp)
plot(harran_k)

harran_ke <- envelope(harran_ppp,fun = "Kest")
plot(harran_ke)

Inhomogeneous Poissonfunction G/F/K

harran_gi <- Ginhom(harran_ppp,lambda = predict(harran_rhohat)) # harran_rhohat needs an bandwidth of 200
plot(harran_gi)


harran_fi <- Finhom(harran_ppp,lambda = predict(harran_rhohat))
plot(harran_fi)

#par(mfrow = c(1,2))
#plot(harran_gi, xlim = c(0,6000))
#plot(harran_g, xlim = c(0,6000))      Gegenüberstellung