sheck at ucar.edu wrote:
> I would like to create a 3d topographic map using lat/lon and
> z(height). I have been scouring the R help pages and have not located
> the package I am looking for. Does anyone have a suggestion of package
> that will work for this?
I have some code for generating some pretty cool 3D topographic maps,
even interactive ones. It's not very difficult to do, really, and much
of the code is actually just for generating nice colours for the map. :-)
I thought it was quite surprising and interesting to see how shallow
the ocean is many place, and how much the depth actually varies.
For instance, the North Sea (between the coast of Norway and the
coast of England) is only about 100 meters deep most places. But
in other places the ocean is several kilometers deep.
Anyway, here's the code:
# First download the SRTM30_Plus data file (about 55 MiB) from
# ftp://topex.ucsd.edu/pub/srtm30_plus/srtm30/data/w020n90.Bathymetry.srtm
# See also http://topex.ucsd.edu/WWW_html/srtm30_plus.html
# Bounding box
minlat=40
maxlat=90
minlong=-20
maxlong=20
# Range of bounding box
longrange=maxlong-minlong
latrange=maxlat-minlat
# Number of columns and rows of data in the bathymetry file
ncols=60 * 2 * longrange
nrows=60 * 2 * latrange
ntot=ncols*nrows # Total number of points -- probably too many to plot ...
steps=10 # ... so we use only every 'steps' point in each
direction
ux=seq(1,ncols,by=steps) # X coordinates (longitude)
uy=seq(1,nrows,by=steps) # Y coordinates (latitude)
dat=expand.grid(x=ux,y=uy) # Create grid
dat=transform(dat,index=(y-1)*ncols+x) # Calculate byte index in file
# Transform indices to real coordinates
dat=transform(dat, long=ggplot2::rescale(x, from=c(1,ncols),
to=c(minlong+1/60/4, maxlong-1/60/4), clip=FALSE),
lat=ggplot2::rescale(y, from=c(1,nrows),
to=c(maxlat-1/60/4, minlat+1/60/4), clip=FALSE))
# Name of file containing the bathymetry data
filename="w020n90.Bathymetry.srtm"
# Read bathymetry data from the file
library(R.utils)
dat$z=readBinFragments(filename, what="integer", size=2, signed=TRUE,
idxs=dat$index,endian="big")
# Create matrix containing the data
dat.mat=matrix(dat$z,ncol=length(ux),byrow=TRUE)
dat.mat=t(dat.mat)[,nrow(dat.mat):1]
# Calculate topographic colours for elevation values
map.colours=function(z, water=c(-5000,0), land=c(0,1500), mountain,...)
{
z.min=min(z)
z.max=max(z)
# Divide the land gradient into two parts
land=c(land[1],mean(land),land[2])
# Number of unique water and land values
nw=length(unique(z[z<0]))
nl=length(unique(z[z>=0]))
# Create the output colour component vectors
h <- s <- v <- rep(1, length(z))
# h-values for water
hvals=c(43/60,31/60)
# Create water colours
if(nw==1) h[z<0]=hvals[2] else if(nw>1)
h[z<0]=approx(water,hvals,xout=z[z<0],rule=2)$y
# h, s and v values for land
hvals=c(4/12,2/12,0/12)
svals=c(1,1,0)
vvals=c(.65,.9,.95)
# Create land colours
if(nl==1)
{ h[z>=0] = hvals[1]; s[z>=0] = svals[1]; v[z>=0] = vvals[1]} else
if(nl>1)
{
h[z>=0]=approx(land,hvals,xout=z[z>=0],rule=2)$y
s[z>=0]=approx(land,svals,xout=z[z>=0],rule=2)$y
v[z>=0]=approx(land,vvals,xout=z[z>=0],rule=2)$y
}
hsv(h,s,v)
}
# Basic contour plot -- not very pretty ...
with(dat, contour(unique(long),rev(unique(lat)),dat.mat))
# Colours make the whole thing look quite pretty ...
ncolours=100
image(unique(dat$long),rev(unique(dat$lat)),
matrix(1:length(dat.mat),nrow=nrow(dat.mat)),
col=map.colours(dat.mat),xlab="Longitude",ylab="Latitude")
# And it looks even prettier in 3D ...
zfacet <- dat.mat[-nrow(dat.mat), -ncol(dat.mat)]
with(dat, persp(unique(long), rev(unique(lat)), dat.mat,
expand=.2, col=map.colours(zfacet), border=NA,
shade=.3, phi=60, theta=0, axes=FALSE))
# How about an *interactive* 3D map ...
# Try holding the left mouse button and dragging
# Now try the mouse wheel (or right button and dragging)
# Finally, try clicking the middle button (or wheel) and dragging
library(rgl)
with(dat, surface3d(unique(long), rev(unique(lat)), dat.mat/1000,
col=map.colours(dat.mat)))
# Note that the elevation values are of course very much exaggerated
# In 'real life', it would look something like this (i.e., completely
flat)
with(dat, surface3d(unique(long), rev(unique(lat)), dat.mat/1000000,
col=map.colours(dat.mat)))
--
Karl Ove Hufthammer