initial: Bias dependent on initial condition

In jonasbhend/biascorrection: Bias Correct Ensemble Forecasts

Bias dependent on initial condition

Description

Mean De-biasing in Dependence of Initial Conditions

Usage

initial(fcst, obs, fcst.out = fcst, span = min(1, 31/nrow(fcst)), ...)

Arguments

fcst	n x m x k array of n lead times, m forecasts, of k ensemble members
obs	n x m matrix of veryfing observations
fcst.out	array of forecast values to which bias correction should be applied (defaults to fcst)
span	the parameter which controls the degree of smoothing (see loess)
...	additional arguments for compatibility with other bias correction methods

Details

This bias correction method assumes that the time-dependent mean bias depends on the initial conditions. The method loosely follows the ideas outlined in Fuckar et al. (2014), but in contrast to their approach, we use the forecast and observed conditions at the first day of the forecast as a proxy for the initial condition. Thereby, individual ensemble members have varying bias correction depending on their respective initial conditions.

Examples

## initialise forcast observation pairs
fcst <- array(rnorm(215*30*51), c(215, 30, 51)) + 
0.5*sin(seq(0,4,length=215)) + 
rep(seq(0,1,length=30), each=215)
obs <- array(rnorm(215*30, mean=2), c(215, 30)) + 
sin(seq(0,4, length=215)) + 
rep(seq(0,3,length=30), each=215)
fc.time <- outer(1:215, 1981:2010, function(x,y) as.Date(paste0(y, '-11-01')) - 1 + x)
fcst.debias <- biascorrection:::initial(fcst[,1:20,], 
obs[,1:20], fcst.out=fcst, fc.time=fc.time, span=0.5)

jonasbhend/biascorrection documentation built on Nov. 11, 2023, 1:16 a.m.