R: Multivariate bias correction (N-pdft)

MBCn {MBC}

R Documentation

Multivariate bias correction (N-pdft)

Description

Multivariate bias correction that matches the multivariate distribution using QDM and the N-dimensional probability density function transform (N-pdft) following Cannon (2018).

Usage

MBCn(o.c, m.c, m.p, iter=30, ratio.seq=rep(FALSE, ncol(o.c)),
     trace=0.05, trace.calc=0.5*trace, jitter.factor=0,
     n.tau=NULL, ratio.max=2, ratio.max.trace=10*trace,
     ties='first', qmap.precalc=FALSE, rot.seq=NULL,
     silent=FALSE, n.escore=0, return.all=FALSE, subsample=NULL,
     pp.type=7)

Arguments

`o.c`	matrix of observed samples during the calibration period.
`m.c`	matrix of model outputs during the calibration period.
`m.p`	matrix of model outputs during the projected period.
`iter`	maximum number of algorithm iterations.
`ratio.seq`	vector of logical values indicating if samples are of a ratio quantity (e.g., precipitation).
`trace`	numeric values indicating thresholds below which values of a ratio quantity (e.g., `ratio=TRUE`) should be considered exact zeros.
`trace.calc`	numeric values of thresholds used internally when handling of exact zeros; defaults to one half of `trace`.
`jitter.factor`	optional strength of jittering to be applied when quantities are quantized.
`n.tau`	number of quantiles used in the quantile mapping; `NULL` equals the length of the `m.p` series.
`ratio.max`	numeric values indicating the maximum proportional changes allowed for ratio quantities below the `ratio.max.trace` threshold.
`ratio.max.trace`	numeric values of trace thresholds used to constrain the proportional change in ratio quantities to `ratio.max`; defaults to ten times `trace`.
`ties`	method used to handle ties when calculating ordinal ranks.
`qmap.precalc`	logical value indicating if `m.c` and `m.p` are outputs from `QDM`.
`rot.seq`	use a supplied list of random rotation matrices. `NULL` generates on the fly.
`silent`	logical value indicating if algorithm progress should be reported.
`n.escore`	number of cases used to calculate the energy distance when monitoring convergence.
`return.all`	logical value indicating whether results from all iterations are returned.
`subsample`	use `subsample` draws of size `n.tau` to calculate initial empirical quantiles; if `NULL`, calculate normally.
`pp.type`	type of plotting position used in `quantile`.

Value

a list of with elements consisting of:

`mhat.c`	matrix of bias corrected `m.c` values for the calibration period.
`mhat.p`	matrix of bias corrected `m.p` values for the projection period.

References

Cannon, A.J., 2018. Multivariate quantile mapping bias correction: An N-dimensional probability density function transform for climate model simulations of multiple variables. Climate Dynamics, 50(1-2):31-49. doi:10.1007/s00382-017-3580-6

Cannon, A.J., S.R. Sobie, and T.Q. Murdock, 2015. Bias correction of simulated precipitation by quantile mapping: How well do methods preserve relative changes in quantiles and extremes? Journal of Climate, 28:6938-6959. doi:10.1175/JCLI-D-14-00754.1

Pitié, F., A.C. Kokaram, and R. Dahyot, 2005. N-dimensional probability density function transfer and its application to color transfer. In Tenth IEEE International Conference on Computer Vision, 2005. ICCV 2005. (Vol. 2, pp. 1434-1439). IEEE.