bfmThresh: BFM With Magnitude Threshold
In bendv/bfmThresh: Experimental alternative to bfmSpatial()

Description Usage Arguments Details Value Author(s) Examples

bfm with automatic magn thresholding and re-monitoring

1
2
3

bfmThresh(x, dates, start, monend = NULL, magnThresh = 0, magnPeriod = 1,
  adjust = 1, verbose = FALSE, formula = response ~ harmon, order = 1,
  returnBFM = FALSE, ...)

`x`	Numeric. Time series vector
`dates`	Time (date) index of `x`
`start`	Numeric. Start of monitoring period. See `bfastmonitor`
`monend`	Numeric. Optional: end of the monitoring period. See `bfmSpatial`
`magnThresh`	Numeric. Threshold for breakpoints. All breakpoints with magnitude < `magnThresh` are ignored, all data `adjust` years preceding this breakpoint are deleted, and monitoring continues with a revised history period.
`magnPeriod`	Numeric. Period of time (in years) from which to compute magnitude (as median of residuals)
`adjust`	Numeric. Delete all observations within this many years preceding a breakpoint, if that breakpoint has a magnitude less than `magnThresh`
`verbose`	Logical. Display error messages (if any)?
`formula`	See `bfastmonitor`
`order`	See `bfastmonitor`
`returnBFM`	Logical. Return a regular `bfastmonitor` object?
`...`	Arguments to pass to `bfastmonitor`

The magnitude is computed as the median of the residuals within the period magnPeriod years after the break date. If this magnitude is greater than magnThresh, all observations preceding the break date by adjust years are deleted (including the observation at the break date), the start of the monitoring period is reset to the break date, the model is re-fit and monitoring continues until the end of the time series is reached, or a breakpoint with magnitude > magnThresh is found.

The default value for magnPeriod is 1, assuming that no follow-up land use or other dynamics take place within a 1-year period.

The amplitude of the fitted 1st order model can be computed by taking sqrt{harmoncos^2 + harmonsin^2}

Numeric vector with output: break date, magnitude, length of history period (in years), R-squared of fitted model, Adjusted R-squared of fitted model, Intercept of fitted model, cosine coefficient(s) (if applicable), sine coefficient(s) (if applicable), error flag (1 means an error was encountered in the try() statement). If returnBFM is FALSE, a list with the above values as first item and a regular BFM object as second item is returned.

Ben DeVries

library(bfmThresh)
data(tura)
s <- getSceneinfo(names(tura))
xy <- c(820692.4, 829564.4)
z <- zoo(tura[cellFromXY(tura, xy)][1, ], s$date)
z <- window(z, start = "1999-01-01")
bfm <- bfmThresh(z, dates = time(z), start = c(2005, 1), plot = TRUE)
print(bfm)
bfm2 <- bfmThresh(z, dates = time(z), start = c(2005, 1), magnThresh = -500, plot = TRUE)
print(bfm2)