Interpolation of change values

[DamienIrving]

My summary of previous iterations of qq-scaling code at CSIRO says that the more recent versions "added completely individual percentile binning (100 of them) with interpolation for change values depending on where the individual daily values fell in-between percentile bins."

The empirical quantile mapping functionality in xclim has an interp option, but an example says "to reduce the risk of sharp change in the adjustment at the interface of the months, interp='linear' can be passed to adjust and the adjustment factors will be interpolated linearly. Ex: the factors for the 1st of May will be the average of those for april and those for may." This doesn't sound like the same type of interpolation.

[DamienIrving]

I'm pretty sure the interp option does in fact do what the previous iteration of the qqscaling code does. e.g.

qm.adjust(ds[args.var], extrapolation='constant', interp='linear')

as opposed to

qm.adjust(ds[args.var], extrapolation='constant', interp='nearest')

When you use "nearest" you can tell which days get put in which quantile bins because they all have the same adjustment factor.