In dbarneche/envPred: Calculate Environmental Predictability for Even and Uneven Time Series Data

library(envPred)

First, we recommend carefully reading the help page for the main function

?env_stats

To calculate the five environmental statistics (seasonality, colour of environmental noise, predictability, constancy, contingency) using sample data (temporally even SST data):

sst_pred <- env_stats(time_series = sst$time_series, dates = sst$dates,
                      n_states = 11, delta = 1, is_uneven = FALSE,
                      interpolate = FALSE, show_warns = TRUE,
                      noise_method = "spectrum")

The argument n_states is a numeric vector of length 1 containing a somewhat arbitrary number, as Colwell's method divides a continuous variable up into discrete states (read the original paper for further details). Default (arbitrary) is 11.

The above example threw a warning, which can be addressed in this example by:

sst  <-  sst[sst$dates <= as.Date("2007-01-01"), ]
sst_pred <- env_stats(sst$time_series, sst$dates, n_states = 11, delta = 1,
                      is_uneven = FALSE, interpolate = FALSE,
                      show_warns = TRUE, noise_method = "spectrum")

The data can be plotted using ggplot2 methods. The user can plot both the de-trended dataset overlaid with the seasonal interpolation,

gg_envpred(sst_pred, type = "detrended")

or the spectral density on the log-10 scale

gg_envpred(sst_pred, type = "spectral")

Finally, the package also handles temporally uneven NPP data with NAs (carefully check warnings), by using the Lomb-Scargle Periodogram from package lomb.

npp_pred  <-  env_stats(npp$time_series, npp$dates, n_states = 11, delta = 8,
                        is_uneven = TRUE, interpolate = TRUE,
                        show_warns = TRUE, noise_method = "lomb_scargle")

dbarneche/envPred documentation built on June 28, 2020, 5:04 p.m.