Smooth Raster Time-series

Smooths pixel-level data in raster time-series and can impute missing (NA) values.

smooth.time.series(x, f = 0.8, smooth.data = FALSE, ...)

Arguments

x	A raster stack/brick or sp object with a @data slot
f	Smoothing parameter (see loess span argument)
smooth.data	(FALSE/TRUE) Smooth all of the data or just impute NA values
...	Additional arguments passed to raster calc (for writing results to disk)

Value

A raster stack or brick pr data.frame object with imputed NA values or smoothed data.

Details

This function uses a LOESS regression to smooth the time-series (using the smooth.data = TRUE argument). If the data is smoothed, it will be replaced by a loess estimate of the time-series (estimated distribution at the pixel-level). The results can dramatically be effected by the choice of the smoothing parameter (f) so caution is warranted and the effect of this parameter tested. Alternately, with smooth.data = FALSE, the function can be used to impute missing pixel data (NA) in raster time-series (stacks/bricks).

See also

loess for details on the loess regression

calc for details on additional (...) arguments

Author

Jeffrey S. Evans <jeffrey_evans@tnc.org>

Examples

if (FALSE) {
 random.raster <- function(r=50, c=50, l=10, min=0, max=1){ 
   do.call(stack, replicate(l, raster(matrix(runif(r*c, min, max),r,c))))
 }
 r <- random.raster()

 #### Smooth time-series using raster stack/brick 
 r.smooth <- smooth.time.series(r, f = 0.6, smooth.data = TRUE)  
 
 #### sp SpatialPixelsDataFrame example
 r <- as(r, "SpatialPixelsDataFrame")
 
 # extract pixel 100 for plotting
 y <- as.numeric(r@data[100,])
 
 # Smooth data
 r@data <- smooth.time.series(r, f = 0.6, smooth.data = TRUE)
   
 # plot results  
 plot(y, type="l")
   lines(as.numeric(r@data[100,]), col="red")
     legend("bottomright", legend=c("original","smoothed"),
         lty=c(1,1), col=c("black","red"))  

 # coerce back to raster stack object  
 r <- stack(r) 

}