grid with random or clustered distribution

########################################################
### simulate landscapes with spatial autocorrelation ###
###             Sarah Goslee 2015-09-09              ###
###     Goslee 2006 PLANT ECOLOGY 187(2):203-212     ###
########################################################

library(gstat)


## parameters
abun <- 0.2
dim1 <- 20
dim2 <- 50


## setup
xy <- expand.grid(seq_len(dim1), seq_len(dim2))
names(xy) <- c("x","y")


## three sample simulations

# no spatial autocorrelation
g.dummy <- gstat(formula = z~x+y, locations = ~x+y, dummy = TRUE, beta
= 0, model = vgm(1,"Nug", 0), nmax = 50)
sim <- predict(g.dummy, newdata = xy, nsim = 1)
random.landscape.000 <- predict(g.dummy, newdata = xy, nsim = 1)
random.landscape.000[,3] <- ifelse(random.landscape.000[,3] >
quantile(random.landscape.000[,3], abun), 0, 1)

# little spatial autocorrelation
g.dummy <- gstat(formula = z~x+y, locations = ~x+y, dummy = TRUE, beta
= 0, model = vgm(1,"Exp", 5), nmax = 50)
random.landscape.005 <- predict(g.dummy, newdata = xy, nsim = 1)
random.landscape.005[,3] <- ifelse(random.landscape.005[,3] >
quantile(random.landscape.005[,3], abun), 0, 1)

# much spatial autocorrelation
g.dummy <- gstat(formula = z~x+y, locations = ~x+y, dummy = TRUE, beta
= 0, model = vgm(1,"Exp", 250), nmax = 50)
sim <- predict(g.dummy, newdata = xy, nsim = 1)
random.landscape.250 <- predict(g.dummy, newdata = xy, nsim = 1)
random.landscape.250[,3] <- ifelse(random.landscape.250[,3] >
quantile(random.landscape.250[,3], abun), 0, 1)


# plot the simulated landscapes
par(mfrow=c(1,3))
image(random.landscape.000, main="Null", xaxt="n", yaxt="n", bty="n",
xlim=c(0,dim1), ylim=c(0, dim2), col=c("lightgray", "darkgray"))
image(random.landscape.005, main="5", xaxt="n", yaxt="n", bty="n",
xlim=c(0,dim1), ylim=c(0, dim2), col=c("lightgray", "blue"))
image(random.landscape.250, main="250", sub=paste("abun =", abun),
xaxt="n", yaxt="n", bty="n", xlim=c(0,dim1), ylim=c(0, dim2),
col=c("lightgray", "darkblue"))

########################################################
###                        end                       ###
########################################################

On Wed, Sep 9, 2015 at 9:27 AM, SH <emptican at gmail.com> wrote:

Hi Sarah,

Thanks for your prompt responding.  The methodology in the publication is
very similar to what I plan to do.  Yes, could you be willing to share

the

code if you don't mind?

Thanks a lot again,

Steve

On Wed, Sep 9, 2015 at 9:11 AM, Sarah Goslee <sarah.goslee at gmail.com>

wrote:

You can use gstat, as in:

https://www.researchgate.net/publication/43279659_Behavior_of_Vegetation_Sampling_Methods_in_the_Presence_of_Spatial_Autocorrelation

If you need more detail, I can dig up the code.

Sarah

On Wed, Sep 9, 2015 at 8:49 AM, SH <emptican at gmail.com> wrote:

Hi R-users,

I hope this is not redundant questions.  I tried to search similar
threads
relevant to my questions but could not find.  Any input would be

greatly

appreciated.

I want to generate grid with binary values (1 or 0) in n1 by n2 (e.g.,
100
by 100 or 200 by 500, etc.) given proportions of 1 and 0 values (e.g.,
1,
5, or 10% of 1 from 100 by 100 grid).  For clustered distributed

grid, I

hope to be able to define cluster size if possible.  Is there a simple
way
to generate random/clustered grids with 1 and 0 values with a
pre-defined proportion?

So far, the function "EVariogram" in the "CompRandFld" package

generates

clustered grid with 1 and 0.  Especially, the example #4 in the
"EVariogram" function description is a kind of what I want. Below is

the

slightly modified code from the original one.  However, the code below
can't control proportion of 1 and 0 values and complicated or I have

no

idea how to do it.  I believe there may be easies ways to
generate random/clustered grids with proportional 1 and 0 values.

Thank you very much in advance,

Steve


library(CompRandFld)
library(RandomFields)

x0 <- seq(1, 50, length.out=50)
y0 <- seq(1, 60, length.out=60)
d <- expand.grid(x=x0, y=y0)
dim(d)
head(d)
x <- d$x
y <- d$y
# Set the model's parameters:
corrmodel <- 'exponential'
mean <- 0
sill <- 1
nugget <- 0
scale <- 3
set.seed(1221)
# Simulation of the Binary-Gaussian random field:
data <- RFsim(x, y, corrmodel="exponential", model="BinaryGauss",

 param=list(mean=mean,sill=sill,scale=scale,nugget=nugget),

              threshold=0)$data
# Empirical lorelogram estimation:
fit <- EVariogram(data, x, y, numbins=20, maxdist=7,

type="lorelogram")

# Results:
plot(fit$centers, fit$variograms, xlab='Distance', ylab="Lorelogram",
     ylim=c(min(fit$variograms), max(fit$variograms)),
     xlim=c(0, max(fit$centers)), pch=20, main="Spatial Lorelogram")
# Plotting
plot(d, type='n')
text(d, label=data)

--
Sarah Goslee
http://www.functionaldiversity.org

--
Sarah Goslee
http://www.stringpage.com
http://www.sarahgoslee.com
http://www.functionaldiversity.org