Hurricane Sandy Land Wind Speed and Kriging

November 28, 2012
By

(This article was first published on Statistical Research » R, and kindly contributed to R-bloggers)

NJ Hurricane Sandy Landfall Data

These data come from the National Climatic Data Center (NCDC).  Using the above link will download all of the data collected by the NCDC on the day of Hurricane Sandy.  The data can also be obtained directly from the source at http://cdo.ncdc.noaa.gov/qclcd/QCLCD.

The purpose of this post is not a discussion on kriging or any of its properties. The purpose is to simply provide a simple R example on kriging and how it can be applied on real data.   This R code uses Hurricane Sandy as an example and predicts the wind speed across the state of New Jersey using the measurement stations that the NCDC uses. It is easy to see how this technique can be applied to other data.

 


wd <- "C:\\nj hurricane data"
setwd(wd)

plot.field.points <- function(locs, data, zlim, xlim, ylim,empty_locs=diag(2),
projection, add=FALSE, legend=TRUE,gray=FALSE,
legend.offset=4, pch=20, cex=4, n.color=64,
xlab="",ylab="",
map.borders=NULL, parameters=NULL, ...)
{
require(fields)
require(maps)

miss<-is.na(data)
data<-data[!miss]
locs<-locs[!miss,]


if (missing(zlim)){
zlim <- range(data)
}

if (missing(xlim)){
xlim <- range(locs[ ,1])
}
if (missing(ylim)){
ylim <- range(locs[ ,2])
}
if (!missing(projection)) {
require(mapproj)
proj <- mapproject(locs[ ,1], locs[ ,2], projection=projection)
locs <- cbind(proj$x, proj$y)
}


plot.range <- zlim
# Which data points are outside the allowed range?
outside.range <- ((data < plot.range[1]) | (data > plot.range[2]))
# A scaled version of the data that's between 1 and n.color
pos.data <- data - plot.range[1]
pos.data <- pos.data / (plot.range[2]-plot.range[1]) * (n.color-1) + 1
# Put in a dummy value if the data is not in range
pos.data[outside.range] <- 1
# Assign a color to each data point
cols <- tim.colors(n.color)[pos.data]
if(gray){
grays<-gray(seq(0.99,0.01,length=n.color)^0.5)
cols <- grays[pos.data]
}

# Make the data points that are outside the range white
cols[outside.range] <- "#FFFFFF"

if (!add) {
if (legend) {
old.par <- par(no.readonly=TRUE)
par(oma=c(0,0,0,legend.offset))
if (missing(projection)) {
plot(locs[ ,1], locs[ ,2], xlab=xlab,ylab=ylab,xlim=xlim, ylim=ylim, type="n", ...)
} else {
map(xlim=xlim, ylim=ylim, projection=projection, type="n", ...)
}
points(locs[ ,1], locs[ ,2], col=cols, pch=pch, cex=cex, ...)
points(empty_locs[ ,1], empty_locs[ ,2], col=0,pch=1,cex=1, ...)
if (!is.null(map.borders)) {
if(!missing(projection)) {
map(map.borders, add=TRUE,
xlim=xlim, ylim=ylim, projection=projection, ...)
} else {
map(map.borders, add=TRUE,
xlim=xlim, ylim=ylim, ...)
}
}
par(old.par)
if(!gray){
quilt.plot(x=locs[ ,1], y=locs[ ,2], z=data, legend.only=TRUE, zlim=zlim)
}
if(gray){
quilt.plot(x=locs[ ,1], y=locs[ ,2], z=data, col=grays,legend.only=TRUE, zlim=zlim)
}
} else {
plot(locs[ ,1], locs[ ,2], xlab=xlab,ylab=ylab, col=cols, pch=pch, cex=pch, ...)
}
} else {
points(locs[ ,1], locs[ ,2], col=cols, cex=cex, pch=pch, ...)
}
}

&nbsp;
my.files <- dir(wd, pattern = ".txt", full.names = TRUE, ignore.case = TRUE)

krig <- matrix(NA, nrow=length(my.files), ncol=3)
for(i in 1:length(my.files)){

raw <- read.csv(my.files[i], skip=6, header=T)
raw <- subset(raw, raw$Date=="20121029")

latlongraw <- read.csv(my.files[i], skip=3, header=F, sep="")
latlongrows <- head(latlongraw,2)[,2]
latlong <- as.numeric(levels(latlongrows)[latlongrows])

raw$WindSpeed[raw$WindSpeed=="M"] <- NA
if(is.factor(raw$WindSpeed)){
raw$WindSpeed <- as.numeric(levels(raw$WindSpeed)[raw$WindSpeed])
}
krig[i,1] <- latlong[1]
krig[i,2] <- latlong[2]
krig[i,3] <- max(raw$WindSpeed, na.rm=T)

}
lat <- krig[,1]
long <- krig[,2]
s<-cbind(long,lat)
PM <- krig[,3]
sandy$coords <- s
sandy$data <- PM

ml <- likfit(sandy,
fix.nugget=F, cov.model="exponential",
ini = c(10, 5), nugget=4)

ml
summary(ml)

grid <- map("state","new jersey", plot=FALSE)
grid.plot <- cbind(
c(min(grid$x, na.rm=TRUE),max(grid$x, na.rm=TRUE)),
c(min(grid$y, na.rm=TRUE),max(grid$y, na.rm=TRUE))
)

sp1<-seq(grid.plot[1,1]-.25,grid.plot[2,1]+.25,length=100)
sp2<-seq(grid.plot[1,2]-.25,grid.plot[2,2]+.25,length=100)
sp<-expand.grid(sp1,sp2)
inLoc<-map.where("state",x=sp[,1],y=sp[,2])
inLoc[is.na(inLoc)]<-"NA"
inLoc<-inLoc=="new jersey"
#Perform ordinary Kriging (value of cov.pars and nugget are copied from mle output):
pred<-krige.conv(data=PM,coords=s,locations=sp,
krige=krige.control(type.krige="ok",cov.model="exponential",
cov.pars=c(100,3),
nugget=5))

pred$predict[!inLoc]<-NA
pred$krige.var[!inLoc]<-NA

#Plot the predicted values:
image.plot(sp1,sp2,matrix(pred$predict,100,100),zlim=range(PM), main="Sandy Maximum Wind Speed in MPH", xlab="Longitude",ylab="Latitude")
map("county",add=T)
points(s, pch=16)

To leave a comment for the author, please follow the link and comment on his blog: Statistical Research » R.

R-bloggers.com offers daily e-mail updates about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...



If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.