trend {cmsafops}

Determine linear trends.

Description

The function determines the trend from data of a single CM SAF NetCDF input file basing on a simple linear model. Depending on the file size, this function could be very time consuming, thus there are two available options. Option 1 (default) is using an apply approach and will read the whole data in once. This option is quite fast, but requires enough memory. Option 2 is using the same calculation, but reads the data pixel by pixel, which is very slow, but can also be applied for large data files, which would not fit into the memory at once.

Usage

trend(
  var,
  infile,
  outfile,
  option = 1,
  nc34 = 4,
  overwrite = FALSE,
  verbose = FALSE,
  nc = NULL
)

Arguments

var

Name of NetCDF variable (character).

infile

Filename of input NetCDF file. This may include the directory (character).

outfile

Filename of output NetCDF file. This may include the directory (character).

option

The way of data handling. Option = 1 is fast but memory consuming (default). Option = 2 is slow, but needs much less memory. Input is either 1 or 2 (numeric).

nc34

NetCDF version of output file. If nc34 = 3 the output file will be in NetCDFv3 format (numeric). Default output is NetCDFv4.

overwrite

logical; should existing output file be overwritten?

verbose

logical; if TRUE, progress messages are shown

nc

Alternatively to infile you can specify the input as an object of class ncdf4 (as returned from ncdf4::nc_open).

Value

A NetCDF file including three data layers is written. One layer (trend1) contains the linear trend multiplied by the number of time steps. In older versions of the package (<= 1.7) the trend was given in the same way as trend1. Another layer (trend2) contains just the calculated linear trend. An additional layer contains a measure for the significance of the calculated trends, which was derived using the 95 % confidence interval. The significance is calculated from the lower and upper value of the 95% confidence interval: lower or upper value < 0: sig = 0 (not significant); lower and upper value < 0: sig = -1 (negative significant); lower and upper value > 0: sig = 1 (positive significant)

See Also

Other temporal operators: cmsaf.detrend(), cmsaf.mk.test(), cmsaf.regres(), num_above(), num_below(), num_equal(), timavg(), timmax(), timmean(), timmin(), timpctl(), timsd(), timsum(), trend_advanced()

Examples

## Create an example NetCDF file with a similar structure as used by CM
## SAF. The file is created with the ncdf4 package.  Alternatively
## example data can be freely downloaded here: <https://wui.cmsaf.eu/>

library(ncdf4)

## create some (non-realistic) example data

lon <- seq(5, 15, 0.5)
lat <- seq(45, 55, 0.5)
time <- seq(as.Date("2000-01-01"), as.Date("2010-12-31"), "month")
origin <- as.Date("1983-01-01 00:00:00")
time <- as.numeric(difftime(time, origin, units = "hour"))
data <- array(250:350, dim = c(21, 21, 132))

## create example NetCDF

x <- ncdim_def(name = "lon", units = "degrees_east", vals = lon)
y <- ncdim_def(name = "lat", units = "degrees_north", vals = lat)
t <- ncdim_def(name = "time", units = "hours since 1983-01-01 00:00:00",
 vals = time, unlim = TRUE)
var1 <- ncvar_def("SIS", "W m-2", list(x, y, t), -1, prec = "short")
vars <- list(var1)
ncnew <- nc_create(file.path(tempdir(),"CMSAF_example_file.nc"), vars)
ncvar_put(ncnew, var1, data)
ncatt_put(ncnew, "lon", "standard_name", "longitude", prec = "text")
ncatt_put(ncnew, "lat", "standard_name", "latitude", prec = "text")
nc_close(ncnew)

## Determine the trend of the example CM SAF NetCDF file and write the
## output to a new file.
trend(var = "SIS", infile = file.path(tempdir(),"CMSAF_example_file.nc"), 
 outfile = file.path(tempdir(),"CMSAF_example_file_trend.nc"))

unlink(c(file.path(tempdir(),"CMSAF_example_file.nc"), 
 file.path(tempdir(),"CMSAF_example_file_trend.nc")))