time series - band filtering based on PSD, to filter out frequency domains in r, probably using "buttord" from signal -

i'm still novice in r , read quite couple of posts , discussions on how filter out frequency domains in time series, none of quite matched problem.

i ask suggestions following: calculated wavelet coherence 2 annually measured time series , taking @ wavelet coherence psd graph:

the purple line (i.e. 8 year period) represents border under filter out frequency domain, not in psd, in original input data.

i though using butter function signal package, overcomplicated purposes.

thus approached problem bwfilter function of mfilter package fo pass through data on 8 year period corresponds 2.37e-7 hz.

name="dta oak.resid tair " adat=read.table(file=paste(name,".csv", sep=""), sep=";", header=t) dta=adat$ya highpass <- bwfilter(dta, freq=8,drift=false) plot(highpass) 

however, results not seem correct, because seems filter out data, trend aligned original time series.

do have idea may have gone wrong? measurement unit maybe?

any appreciated , if additional details needed happy provide them! thank you!

the data can found here