climdex.tx90p: Computation of these percentiles involves use of a boostrap...

In climdex.pcic: PCIC Implementation of Climdex Routines

Description

Computation of these values outside of the base period involves comparing the temperature data for each day with the corresponding percentiles for a 5 day running window surrounding that day. The resulting monthly series is then the monthly percentage of values that meet the criteria.

Usage

climdex.tx90p(ci, freq = c("monthly", "annual"))

Arguments

ci	Object of type climdexInput.
freq	Time frequency to aggregate to.

Details

Computation of these values inside the base period is more complicated. It involves comparison of the daily temperature data with the corresponding day of temperature data in each of (n - 1) sets of data. The sets consist of the data for the base period with the current year replaced with each of the other years. The results of these comparisons are then averaged to give a value between 0 and 1. Finally, the resulting daily series is aggregated to a monthly series by averaging these daily values and multiplying by 100 to give a monthly percentile value.

Value

A vector containing a timeseries containing values of the index on a monthly or annual timescale.

Note

These functions may calculate slightly different results than fclimdex.

The bootstrapping method is not well defined for cases where the base data contains numerous missing values. Because of that, this code (and fclimdex) are not very robust against missing values with respect to these indicies. When computing percentiles inside the base period, both this implementation and fclimdex do not divide through by the number of non-missing values when aggregating the values inside the base period. Instead, they divide through by the number of base years minus one. This will result in a negative bias when missing values are present.

References

http://etccdi.pacificclimate.org/list_27_indices.shtml

See Also

climdexInput.raw, climdexInput.csv.

Examples

library(PCICt)

## Create a climdexInput object from some data already loaded in and
## ready to go.

## Parse the dates into PCICt.
tmax.dates <- as.PCICt(do.call(paste, ec.1018935.tmax[,c("year",
"jday")]), format="%Y %j", cal="gregorian")
tmin.dates <- as.PCICt(do.call(paste, ec.1018935.tmin[,c("year",
"jday")]), format="%Y %j", cal="gregorian")
prec.dates <- as.PCICt(do.call(paste, ec.1018935.prec[,c("year",
"jday")]), format="%Y %j", cal="gregorian")

## Load the data in.
ci <- climdexInput.raw(ec.1018935.tmax$MAX_TEMP,
ec.1018935.tmin$MIN_TEMP, ec.1018935.prec$ONE_DAY_PRECIPITATION,
tmax.dates, tmin.dates, prec.dates, base.range=c(1971, 2000))

## Create a monthly timeseries of the TX90p index.
tx90p <- climdex.tx90p(ci)

Example output

Loading required package: PCICt

climdex.pcic documentation built on March 26, 2020, 7:17 p.m.