oce (version 0.9-17)

interpBarnes: Grid data using Barnes algorithm

Description

Grid data using Barnes algorithm.

Usage

interpBarnes(x, y, z, w,
    xg, yg, xgl, ygl,
    xr, yr, gamma=0.5, iterations=2, trim=0,
    debug=getOption("oceDebug"))

Arguments

x, y

a vector of x and ylocations.

z

a vector of z values, one at each (x,y) location.

w

a optional vector of weights at the (x,y) location. If not supplied, then a weight of 1 is used for each point, which means equal weighting. Higher weights give data points more influence.

xg, yg

optional vectors defining the x and y grids. If not supplied, these values are inferred from the data, using e.g. pretty(x, n=50).

xgl, ygl

optional lengths of the x and y grids, to be constructed with seq spanning the data range. These values xgl are only examined if xg and yg are not supplied.

xr,yr

optional values defining the width of the radius ellipse in the x and y directions. If not supplied, these are calculated as the span of x and y over the square root of the number of data.

gamma

grid-focussing parameter. At each iteration, xr and yr are reduced by a factor of sqrt(gamma).

iterations

number of iterations.

trim

a number between 0 and 1, indicating the quantile of data weight to be used as a criterion for blanking out the gridded value (using NA). If 0, the whole zg grid is returned. If >0, any spots on the grid where

debug

a flag that turns on debugging. Set to 0 for no debugging information, to 1 for more, etc; the value is reduced by 1 for each descendent function call.

Value

• A list containing: xg, a vector holding the x-grid); yg, a vector holding the y-grid; zg, a matrix holding the gridded values; wg, a matrix holding the weights used in the interpolation at its final iteration; and zd, a vector of the same length as x, which holds the interpolated values at the data points.

concept

tide

Details

The algorithm follows that described by Koch et al. (1983), with the addition of the ability to blank out the grid in spots where data are sparse, using the trim argument.

References

S. E. Koch and M. DesJardins and P. J. Kocin, 1983. ``An interactive Barnes objective map anlaysis scheme for use with satellite and conventional data,'' J. Climate Appl. Met., vol 22, p. 1487-1503.

See Also

See wind.

Examples

library(oce)

# 1. contouring example, with wind-speed data from Koch et al. (1983)
data(wind)
u <- interpBarnes(wind$x, wind$y, wind$z)
contour(u$xg, u$yg, u$zg, labcex=1)
text(wind$x, wind$y, wind$z, cex=0.7, col="blue")
title("Numbers are the data")

# 2. As 1, but blank out spots where data are sparse
u <- interpBarnes(wind$x, wind$y, wind$z, trim=0.1)
contour(u$xg, u$yg, u$zg, level=seq(0, 30, 1)) 
points(wind$x, wind$y, cex=1.5, pch=20, col="blue")

# 3. As 1, but interpolate back to points, and display the percent mismatch
u <- interpBarnes(wind$x, wind$y, wind$z)
contour(u$xg, u$yg, u$zg, labcex=1)
mismatch <- 100 * (wind$z - u$zd) / wind$z
text(wind$x, wind$y, round(mismatch), col="blue")
title("Numbers are percent mismatch between grid and data")


# 4. As 3, but contour the mismatch
mismatchGrid <- interpBarnes(wind$x, wind$y, mismatch)
contour(mismatchGrid$xg, mismatchGrid$yg, mismatchGrid$zg, labcex=1)

# 5. One-dimensional example, smoothing a salinity profile
data(ctd)
p <- pressure(ctd)
y <- rep(1, length(p)) # fake y data, with arbitrary value
S <- salinity(ctd)
pg <- pretty(p, n=100)
g <- interpBarnes(p, y, S, xg=pg, xr=1)
plot(S, p, cex=0.5, col="blue", ylim=rev(range(p)))
lines(g$zg, g$xg, col="red")