burnr (version 0.2.0)

sea: Perform superposed epoch analysis.

Description

Perform superposed epoch analysis.

Usage

sea(x, event, nbefore = 6, nafter = 4, event_range = TRUE,
  n_iter = 1000)

Arguments

x

A data.frame climate reconstruction or tree-ring series with row names as years.

event

A vector of event years for superposed epoch, such as fire years, or an fhx object with a single series as produced by composite

nbefore

The number of lag years prior to the event year

nafter

The number of lag years following the event year

event_range

Logical. Constrains the time series to the time period of key events within the range of the x climate series. False uses the entire climate series, ignoring the period of key events. time series

n_iter

The number of iterations for bootstrap resampling

Value

A list of three data frames, following the output of EVENT. (1) the actual events table, (2) the simulated events table, and (3) departures of actual from simulated

Details

Superposed epoch analysis (SEA) helps to evaluate fire-climate relationships in studies of tree-ring fire history. It works by compositing the values of an anual timeseries or climate reconstruction for the fire years provided (key) and both positive and negative lag years. Bootstrap resampling of the timeseries is performed to evaluate the statistical significance of each year's mean value. Users interpret the departure of the actual event year means from the simulated event year means.

The significance of lag-year departures from the average climate condition was first noted by Baisan and Swetnam (1990) and used in an organized SEA by Swetnam (1993). Since then, the procedure has been commonly applied in fire history studies. The FORTRAN program EVENT.exe was written by Richard Holmes and Thomas Swetnam (Holmes and Swetnam 1994) to perform SEA for fire history specifically. EVENT was incorporated in the FHX2 software by Henri Grissino-Mayer.

sea was designed to replicate EVENT as closely as possible. We have tried to stay true to their implementation of SEA, although multiple versions of the analysis exist in the climate literature and for fire history (e.g., FHAES implements a diferent procedure). The outcome of EVENT and sea should only differ slightly in the values of the simulated events and the departures, because random draws are used. The event year and lag significance levels should match, at least in the general pattern.

We note that our implementation of run_sea borrows from the dplR::sea function in how it performs the bootstrap procedure, but differs in the kind of output provided for the user.

References

Baisan and Swetnam 1990, Fire history on desert mountain range: Rincon Mountain Wilderness, Arizona, U.S.A. Canadian Journal of Forest Research 20:1559-1569.

Bunn 2008, A dendrochronology program library in R (dplR), Dendrochronologia 26:115-124

Holmes and Swetnam 1994, EVENT program desription

Swetnam 1993, Fire history and climate change in giant sequoia groves, Science 262:885-889.

Examples

## Not run: ------------------------------------
# # Read in the Cook and Krusic (2004; The North American Drought Atlas) reconstruction
# # of Palmer Drought Severity Index (PDSI) for the Jemez Mountains area (gridpoint 133).
## ---------------------------------------------
## Not run: ------------------------------------
# # For users who want to perform SEA very near to EVENT.exe and/or have reproducable draws from
# # the bootstrap procedure, consider including the \code{set.seed} function prior to \code{run_sea}.
# # Convention is to provide a long integer, such as a birthday (e.g. 3191982).
# # In the EVENT.exe program, Richard Holmes used the number of days since 1 January 1935.
# days <- as.numeric(Sys.Date() - as.Date("1jan1935", "%d%b%Y"))
# set.seed(days)
## ---------------------------------------------