Temporal smoothing and gap filling using phenopix

Description

Time series smoothing and gap filling using fitting methods as provided in the greenProcess function of the phenopix package. Function fits are performed for each year separately for which PhenopixMY is used.

Usage

TSGFphenopix(Yt, interpolate = FALSE, fit = "spline", ...)

Arguments

Value

The function returns a gap-filled and smoothed version of the time series.

Author(s)

Matthias Forkel <matthias.forkel@geo.tuwien.ac.at> [aut, cre]

See Also

PhenopixMY, TsPP

Examples

# load a time series of NDVI (normalized difference vegetation index)
data(ndvi)
plot(ndvi)

# introduce random gaps 
gaps <- ndvi
gaps[runif(100, 1, length(ndvi))] <- NA
plot(gaps)

# do smoothing and gap filling
tsgf <- TSGFphenopix(gaps, fit="spline")
plot(gaps)
lines(tsgf, col="red")

# compare original data with gap-filled data
plot(ndvi[is.na(gaps)], tsgf[is.na(gaps)], xlab="original", ylab="gap filled")
abline(0,1)
r <- cor(ndvi[is.na(gaps)], tsgf[is.na(gaps)])
legend("topleft", paste("Cor =", round(r, 3)))

# compare spline from phenopix with TSGFspline
spl <- TSGFspline(gaps)
plot(gaps)
lines(tsgf, col="red")
lines(spl, col="blue")
legend("topleft", c("TSGFphenopix.spline", "TSGFspline"), text.col=c("red", "blue"))
# Note that the differences originate from the fact that TSGFspline is applied on 
# the full time series whereas spline within phenopix is applied for each year 
# separetely. Yearly fits for TSGFphenopix.spline are afterwards combined to a full 
# time series. This can cause jumps or peaks between two years. Thus, TSGFspline is 
# the better choice for multi-year time series. This is also seen in cross-validation:
plot(ndvi[is.na(gaps)], tsgf[is.na(gaps)], xlab="original", ylab="gap filled", col="red")
points(ndvi[is.na(gaps)], spl[is.na(gaps)], col="blue")
abline(0,1)
r <- cor(cbind(ndvi[is.na(gaps)], tsgf[is.na(gaps)], spl[is.na(gaps)]))
lgd <- paste(c("TSGFphenopix.spline", "TSGFspline"), "Cor =", round(r[1,2:3], 3))
legend("topleft", lgd, text.col=c("red", "blue"))

# Other fits wihtin phenopix might be usefull but are rather computationally expensive:
tsgf <- TSGFphenopix(gaps, fit="klosterman")
plot(gaps)
lines(tsgf, col="red")