Global Mann-Kendall trend test

The function significantTau has been developed and added to gimms to facilitate the calculation of reliable long-term monotonous trends. It is up to the user to decide whether or not to apply pre-whitening prior to the Mann-Kendall trend test in order to account for lag-1 autocorrelation. Currently, the function supports the pre-whitening algorithms proposed by Yue, Pilon, Phinney, et al. (2002) and Zhang, Vincent, Hogg, et al. (2000) which are both included in the zyp package (Bronaugh and Werner, 2013). If no pre-whitening is desired, significantTau is merely a wrapper function around MannKendall from Kendall (McLeod, 2011).

### download data -----

## number of cores for parallel processing
cores <- parallel::detectCores() - 1

## download entire gimms ndvi3g collection in parallel
gimms_fls <- downloadGimms(x = as.Date("1982-01-01"), cores = cores)


### rasterize images including quality control -----

## reference extent
library(rworldmap)
deu <- subset(countriesCoarse, ADMIN == "Germany")

## rasterize
gimms_rst <- rasterizeGimms(gimms_fls, ext = deu, keep = 0, cores = cores,
                            filename = gsub(".nc4", ".tif", gimms_files),
                            format = "GTiff", overwrite = TRUE)


### remove seasonal signal -----

library(remote)
gimms_dsn <- deseason(gimms_rst, cycle.window = 24L)


### mann-kendall trend test (p < 0.001) -----

gimms_mk <- significantTau(gimms_dsn, prewhitening = TRUE)
gimms_mk

Figure 3. Long-term Mann-Kendall trends (1982 to 2015; p < 0.001) over Germany derived from GIMMS NDVI3g.v1.