| TrendUncertainty {greenbrown} | R Documentation |
The function computes trend statistics (linear trend slope and intercept, Mann-Kendall tau and p-value) with associated uncertainties (standard deviation) by sampling the time series according to different start and end dates using the function TrendSample
TrendUncertainty(Yt, seg = NULL, bp = NoBP(), sample.method = c("sample",
"all", "none"), sample.min.length = 0.75, sample.size = 30,
fun.unc = NULL, trend = TrendAAT)
Yt |
univariate time series of class |
seg |
a vector indicating segments of a time series. If NULL, provide bp |
bp |
detected breakpoints in the time series as returned by |
sample.method |
Sampling method for combinations of start and end dates to compute uncertainties in trends. If "sample" (default), trend statistics are computed for a sample of combinations of start and end dates according to |
sample.min.length |
Minimum length of the time series (as a fraction of total length) that should be used to compute trend statistics. Time windows between start and end that are shorter than min.length will be not used for trend computation. |
sample.size |
sample size (number of combinations of start and end dates) to be used if |
fun.unc |
function to summarize the uncertainty of the trend (default: quantile 0.025 and 0.975). Can be also 'sd' or other functions. |
trend |
method that should be used to compute the trend |
The function returns a data.frame with the estimated Mann-Kendall tau, p-value and slope and intercept of a linear trend with uncertainties defined as the standard deviation of these estimates dependent on different start and end dates.
Matthias Forkel <matthias.forkel@geo.tuwien.ac.at> [aut, cre]
# load a time series of NDVI (normalized difference vegetation index) data(ndvi) # aggregate time series to annual time steps ndvi <- aggregate(ndvi, FUN=mean, na.rm=TRUE) plot(ndvi) # compute trend statistics dependent on start and end of the time series trd.ens <- TrendSample(ndvi) plot(trd.ens) # compute statistics for trend TrendUncertainty(ndvi) # compute trend statistics with uncertainties by considering breakpoints bp <- breakpoints(ndvi ~ time(ndvi)) trd.unc <- TrendUncertainty(ndvi, bp=bp) trd.unc trd.unc[[1]]$slope_unc