preparing data for weather regimes

[bird-house]

preparing data for weather regimes

[nilshempelmann]

subtraction of seasonal mean requires a smoothing:

"sousseasmean"<-function(dat,conv.time,l.mon=1:12,
l.year=unique(conv.time$year),rprint=FALSE)
{
dat.m=dat
seas.cyc=c()
time.cyc=c()

Calcul du cycle saisonnier pour une liste d'annees dans l.year

if(rprint) print("Seasonal cycle")
for(mon in l.mon){
r.day=range(conv.time$day[conv.time$month==mon])
for(day in r.day[1]:r.day[2]){
m.day=apply(dat[conv.time$day==day & conv.time$month==mon &
conv.time$year %in% l.year,],2,mean,na.rm=TRUE)
seas.cyc=rbind(seas.cyc,m.day)
time.cyc$month=c(time.cyc$month,mon)
time.cyc$day=c(time.cyc$day,day)
}
}

Lissage par spline du cycle saisonnier avec preservation des derivees aux bords

if(rprint) print("Seasonal cycle smoothing")
seas.cyc.spl=rbind(seas.cyc,seas.cyc,seas.cyc)
for(i in 1:ncol(seas.cyc)){
seas.cyc.spl[,i]=smooth.spline(seas.cyc.spl[,i],spar=0.8)$y
}
seas.cyc.spl=seas.cyc.spl[(nrow(seas.cyc)+1):(2*nrow(seas.cyc)),]

Soustraction du cycle saisonnier

if(rprint) print("Seasonal anomalies")
for(t in 1:nrow(dat)){
mon=conv.time$month[t]
day=conv.time$day[t]
ii= which(time.cyc$month %in% mon & time.cyc$day %in% day)
dat.m[t,]=dat[t,]-seas.cyc.spl[ii,]
}
datsub=list(anom=dat.m,seascyc=list(seascyc=seas.cyc.spl,timecyc=time.cyc))
}# fin de la definition de fonction

[nilshempelmann]

That's the R - smooth ....

[nilshempelmann]

a repoduction with python.

[nilshempelmann]

R routine is not handling different calenders:
CMIP5 dataset exists with with 365 days (no 29th feb) calenders.
HadGEM is comming with 360 days

but 'sousseasmean' expect 366 days for the annual cycle

[mcalvcas]

[nilshempelmann]

Done ;-)