baytsDDSpatial: Function to run baytsDD on (mulitple) raster bricks

In jreiche/bayts: Probabalistic approach to combine multiple time series and Bayesian probability updating for near real-time forest change detection

Description

Implements baytsDD function on (multiple) time series rasterBrick object(s).

Usage

baytsDDSpatial(bL = list(), datesL = list(), pdfsdL = list(),
  distNFL = list(), modL = list(), formulaL = list(), orderL = list(),
  mask = NULL, start_history = NULL, end_history = NULL, start,
  end = NULL, chi = 0.9, PNFmin = 0.5, bwf = c(0.1, 0.9),
  mc.cores = 1, out_file = NULL)

Arguments

bL	list of raster bricks. Raster bricks need to have the same extent and spatial resolution.
datesL	list of time vector of the format: "2014-10-07".
pdfsdL	list of pdfsd object(s) describing the modulation of the sd of F and NF sd(F),mean(NF),sd(NF) (e.g. pdfsd = c(2,-4,2))
modL	list of modL - modulation of the time series observations. default=NULL
formulaL	list of formula for the regression model. The default is response ~ trend + harmon, i.e., a linear trend and a harmonic season component. Other specifications are possible using all terms set up by bfastpp, i.e., season (seasonal pattern with dummy variables), lag (autoregressive terms), slag (seasonal autoregressive terms), or xreg (further covariates). See bfastpp for details.
orderL	list of numeric. Order of the harmonic term, defaulting to 3.
mask	(raster) mask at which method is applied; default = NULL (method is applied to all pixel)
start_history	Start date of history period used to model the seasonality and derive F and NF PDFs. Default=NULL (start of input time series)
end_history	End date of history period used to model the seasonality and derive F and NF PDFs. Default=NULL (Start of the monitoring period is used)
start	start date of monitoring period. Default=NULL (start of input time series).
end	end date of monitoring period. Default=NULL (end of input time series)
chi	threshold of Pchange at which the change is confirmed; Default=0.5
PNFmin	threshold of pNF above which the first observation is flagged; Default=0.5
bwf	block weighting function to truncate the NF probability; Default=c(0.1,0.9); (c(0,1) = no truncation)
mc.cores	numeric. number of cores to be used for the job. See mc.calc for more details (default = 1)
distL	list of "distNF" object(s) describing the mean and sd of the NF distribution in case no data driven way to derive the NF distribution is wanted; default=NULL
outfile	output file

Value

A rasterBrick with 5 layers: (1) flag: time at which unconfirmed change got flagged; (2) change.flagged: time at which confirmed change got flagged; (3) change.confirmed: time at which change is confirmed; (4) Pflag: Probabilty of change for unconfirmed flagged changes; (5) Pchange.confirmed: Probabilty of change for confirmed changes.

Author(s)

Johannes Reiche (Wageningen University)

References

Reiche et al. (2015): A Bayesian Approach to Combine Landsat and ALOS PALSAR Time Series for Near Real-Time Deforestation Detection. Remote Sensing. 7(5), 4973-4996; doi:10.3390/rs70504973

Examples

#TBD

jreiche/bayts documentation built on Feb. 3, 2021, 1:12 a.m.