Introduction to dfoliatR

Chris Guiterman

2020-09-02

Installation

dfoliatR is not currently on CRAN. To install dfoliatR use the devtools package:

library(devtools)

install_github("chguiterman/dfoliatR")

Once installed, dfoliatR can be called like any other R package

library(dfoliatR)

Data

dfoliatR requires two independent datasets:

Both host and non-host data should be formatted in R to match the dplR::rwl object type.

Data are provided in dfoliatR to demonstrate some of the utilities of dfoliatR, courtesy of Dr. Ann Lynch.

The first is a collection of Douglas-fir trees sampled for a reconstruction of western spruce budworm. It is paired with a ponderosa pine non-host chronology from a nearby site.

data(dmj_h)

The host and non-host objects are directly from dplR in the form of rwl objects:

DMJ01 DMJ02 DMJ03 DMJ04 DMJ06 DMJ07 DMJ11 DMJ14 DMJ15 DMJ16
1620 NA NA NA NA NA NA NA 1.14221 NA NA
1621 NA NA NA NA NA NA NA 0.84523 NA NA
1622 NA NA NA NA NA NA NA 0.39534 NA NA
1623 NA NA NA NA NA NA NA 0.64625 NA NA
1624 NA NA NA NA NA NA NA 0.67192 NA NA
1625 NA NA NA NA NA NA NA 0.61729 NA NA
1626 NA NA NA NA NA NA NA 0.94704 NA NA
1627 NA NA NA NA NA NA NA 1.21904 NA NA
1628 NA NA NA NA NA NA NA 0.99266 NA NA
1629 NA NA NA NA NA NA NA 1.29200 NA NA
data(dmj_nh)
WIR S
1675 0.537
1676 0.780
1677 1.188
1678 1.496
1679 0.781
1680 1.292
1681 1.256
1682 1.117
1683 1.840
1684 0.989

The dfoliatR package also includes a second site for users to explore:

data(ef_h)

data(ef_nh)

All data created and presented in this vignette is available through the package by the object names used here.

Performance

Analyzing insect outbreak signals from tree rings in dfoliatR requires a two-step process.

  1. Identify defoliation events on host trees by removing the growth pattern of non-host trees. Host and non-host trees are presumed to respond similarly to climate, so this step produces individual tree-level series of what the two species do NOT share in common.

  2. Composite host individual tree series to the site level to assess the scale of defoliation. Events recorded on more than a threshold number and/or percentage of trees (e.g., 25%) are considered outbreaks.

To identify defoliation events, input the set of host tree series and the non-host chronology into the defoliate_trees() function. Some parameters regarding the length and severity of growth departure can be changed by the user. The parameter defaults follow those in OUTBREAK (negative departures for 8 or more years, at least one reaching -1.28 standard deviations, and allowing 1 positive excursion before and after the greatest departure). Definitions of the function parameters are provided with ?defoliate_trees

dmj_defol <- defoliate_trees(host_tree = dmj_h, nonhost_chron = dmj_nh, 
      duration_years = 8, max_reduction = -1.28, list_output = FALSE)
year series gsi ngsi defol_status
1810 DMJ01 1.0830 0.3084 nd
1811 DMJ01 1.0261 0.1024 nd
1812 DMJ01 1.0971 0.3595 nd
1813 DMJ01 0.9828 -0.0546 nd
1814 DMJ01 1.0339 0.1307 nd
1815 DMJ01 1.3461 1.2616 nd
1816 DMJ01 1.4440 1.6166 nd
1817 DMJ01 0.9790 -0.0682 nd
1818 DMJ01 0.7929 -0.7424 nd
1819 DMJ01 0.9078 -0.3264 nd

The best way to evaluate the results of the call to defoliate_trees is to graph the resulting “defol” object:

plot(dmj_defol)

Each horizontal line in the plot provides the measured time sequence for each tree. Defoliation events are shown as thicker line segments, with colors to indicate the relative severity of each event. Breakpoints to distinguish between severe, moderate, and minor defoliation levels can be set via the breaks parameter in plot or plot_defol.

Basic and informative tree-level statistics regarding the sample data and defoliation events are provided.

defol_stats(dmj_defol)
series first last years n_events tot_years mean_duration
DMJ01 1810 1996 187 4 50 12
DMJ02 1750 1996 247 6 75 12
DMJ03 1830 1996 167 4 42 10
DMJ04 1720 1996 277 8 90 11
DMJ06 1700 1996 297 6 81 14
DMJ07 1710 1996 287 7 96 14
DMJ11 1900 1997 98 2 19 10
DMJ14 1675 1996 322 9 112 12
DMJ15 1730 1996 267 4 64 16
DMJ16 1746 1996 251 7 86 12

It is important to note that dfoliatR distinguishes between a “defoliation event” that is recorded on individual trees and an “outbreak” that synchronously effected a proportion of trees.

Outbreak periods can be identified with the function outbreak(). In essence, this is a composite function that combines all trees provided in the “defol” object to assess the synchrony and scale of defoliation. Should enough trees record defoliation (regardless of the duration), it will be termed an “outbreak”. Filtering parameters control the percent of trees in defoliation and minimum number of trees required to be considered an outbreak.

dmj_obr <- outbreak(dmj_defol, filter_perc = 25, filter_min_series = 3)

Running outbreak produces a new class of data frame, an “outbreak” object.

year samp_depth num_defol perc_defol num_max_defol perc_max_defol mean_gsi mean_ngsi outbreak_status
1675 2 0 0.0 0 0.0 1.5303 1.5145 not_obr
1676 2 0 0.0 0 0.0 1.2435 0.7143 not_obr
1677 2 1 50.0 0 0.0 0.9222 -0.1820 not_obr
1678 2 1 50.0 0 0.0 0.9716 -0.0442 not_obr
1679 2 1 50.0 0 0.0 1.1373 0.4198 not_obr
1680 3 1 33.3 0 0.0 0.7244 -0.7303 outbreak
1681 3 1 33.3 0 0.0 0.5545 -1.2048 outbreak
1682 3 1 33.3 0 0.0 0.7372 -0.6983 outbreak
1683 3 2 66.7 1 33.3 0.3466 -1.7847 outbreak
1684 3 2 66.7 0 0.0 0.6063 -1.0589 outbreak

As with “defol” objects, it can be plotted directly, and summary statistics are available.

plot_outbreak(dmj_obr)

dmj_obr_stats <- outbreak_stats(dmj_obr)
head(dmj_obr_stats)
start end duration n_df_start perc_df_start max_df_obr yr_max_df yr_min_ngsi min_gsi min_ngsi
1680 1699 20 1 33.3 3 1690 1692 0.121 -2.407
1753 1769 17 6 46.2 7 1754 1755 0.343 -1.611
1825 1840 16 11 78.6 12 1831 1826 0.500 -1.304
1849 1865 17 7 46.7 13 1852 1853 0.252 -1.994
1881 1895 15 8 53.3 14 1886 1885 0.262 -1.945
1959 1970 12 7 41.2 15 1960 1965 0.328 -1.830
1987 1996 10 9 52.9 15 1989 1995 0.378 -1.640

The summary statistics for “outbreak” objects include each identified outbreak event as a row, with start/end years, duration, and other metrics used for analyses.

One important metric to analyze is the return interval of outbreaks. Many researchers use the first year of the event as the point of reference. One can calculate return intervals for our DMJ site in this way via


dmj_interv <- diff(dmj_obr_stats$start)

# All intervals
dmj_interv
#> [1] 73 72 24 32 78 28

# Mean interval
mean(dmj_interv)
#> [1] 51.16667

# Median interval
median(dmj_interv)
#> [1] 52