write.sst.field {kfsst} | R Documentation |
Staring from a directory (datadir
) of data files
each containing a grid of latitude, longitude, and
corresponding SST measurements, this function reads the
files one by one and spatially smooths the SST-field,
computes the gradient of the smooth field, and write these
informations to a combined file (filename
).
The original data files each represent the SST-field at a
specific time interval (say a week). The time information
is contained in the name of each file.
write.sst.field(datadir, nlon = 100, nlat = 150, filename = "sst.dat", alpha = 0.05, from.ystr = c(3, 6), from.dstr = c(7, 9), to.ystr = c(11, 14), to.dstr = c(15, 17), peak = FALSE)
datadir |
Is the folder where the raw data files are located |
nlon |
The smoothed SST-fields and their gradient-fields are
represented on a nlon -by-nlat grid |
nlat |
The smoothed SST-fields and their gradient-fields are
represented on a nlon -by-nlat grid |
filename |
Is the name of the file where the smoothed fields are saved |
alpha |
Is a scalar between 0 and 1 determining the degree of smoothing used. The default is 5%, which means that the smoothed field at any point is calculated from the 5% nearest points in the observed data. |
from.ystr |
Is an integer vector with two elements describing
what part of the file name describe the year of the
first date the data file represents. For instance if
the names of the data files all have the format
RSyyyyddd_YYYYDDD.dat , where yyyy is
the year of the first date the argument should be
c(3,6) . |
from.dstr |
Is an integer vector with two elements describing what part of the file name describe the 'number of days into the year' of the first date the data file represents. |
to.ystr |
Is similar to from.ystr , but here for the year
of the last date the data file represents. |
to.dstr |
Is similar to from.dstr , but here for the 'number
of days into the year' of the last date the data file
represents. |
peak |
If TRUE allows to visually compare the raw and the
smoothed field. |
The grid size of the internal representation can be finer or coarser than the actual data set, and should be chosen based on size of the area. This way of representing the SST-field is clearly sub-optimal, and will hopefully be replaced in later versions.
The default smoothing scale of 5% is probably coarse in many cases, especially if the area is large.
It is recommended to carry out sensitive analysis with respect to the degree of smoothing and the grid size of the internal representation.
The smoothing is presently done via the locfit
R-package.
The filename returned from the function is where the internal representation is saved.
Anders Nielsen anielsen@dina.kvl.dk, John Sibert sibert@hawaii.edu
# No example supplied here, but check out the example # in the blue.shark dataset documentation