====== Installing ncdf in R ======

**NB** a prerequisite is that the netcdf libraries are already installed.

<code>
  $ mkdir -p ~/.R/libs
  $ export R_LIBS=~/.R/libs
  $ echo 'R_LIBS=~/.R/libs/'
  $ export R_LIBS=~/.R/libs/
  $ echo 'R_LIBS=~/.R/libs/'
  $ echo 'R_LIBS=~/.R/libs/' >> ~/.Renviron
  #Download the ncdf package from CRAN
  #All this can be done in a temporary area ... (?)
  $ tar xvzf ncdf_1.6.tar.gz 
  $ R CMD INSTALL ncdf
  $ history
  ...
  ...
  # there are perhaps even easier ways to do this:

  $ install.packages("gplots", dependencies = TRUE)

  # no source was linked ... but the mirror choose automatically opens
  # and one just selects a site and the install happens!
  ...
  $ 
</code>

Making use of ncdf in R

<code>
  $ R
   Method='unspecified'
   Shortname='outfile'
   library(ncdf)
   nd = open.ncdf("interpolations.nc")
   x=get.var.ncdf(nd,"Original")
   y=get.var.ncdf(nd,"Interpolations")
   x[x==-1]=0
   v <- array(x, length(x))
   w <- array(y, length(y))
   xx=x[!is.na(x) & x>=0 & x < 100]
   yy=y[!is.na(x) & x>=0 & x < 100]
   yyy=yy[!is.na(yy) & yy>=0 & yy < 100]
   xxx=xx[!is.na(yy) & yy>=0 & yy < 100]
   jpeg(filename=paste(Shortname,'jpg',sep='.'))
   plot(xxx,yyy,pch=46,xlab='Original Value',ylab='Interpoltion',main=Method)
   XC=cor(xxx,yyy)
   ff <- lm(xxx ~ yyy)
   ff$coefficients[1]
   ff$coefficients[2]
   z=ff$coefficients[1]*xxx + ff$coefficients[2]
   #lines(xxx,z)
   text(0,95,"Linear Regression",pos=4)
   text(0,90,paste("Gradient =",format(ff$coefficients[1],digits=5)),pos=4)
   text(0,85,paste("Intercept =",format(ff$coefficients[2],digits=5)),pos=4)
   text(0,75,paste("Correlation coefficient =",format(XC,digits=5)),pos=4)
   dev.off()

</code>

Alternatively, place the code in a file and run from the command line:
<code>
 $ Rscript plot_netcdf.R "Test Plot" OUTPUT
</code>
that will produce the file OUTPUT.jpg with the plot title "Test Plot".
  * {{kvalobs:kvoss:system:qc2:8hdk376snf09zj37dk82s92:plot_netcdf.pdf|plot_netcdf.R}}