EOF: Empirical Orthogonal Functions (EOFs).

In metno/esd.test: Climate analysis and empirical-statistical downscaling (ESD) package for monthly and daily data.

Description

Computes EOFs (a type of principal component analysis) for combinations of data sets, typically from gridded data, reanalysis and climate models results.

[ref: Benestad (2001), "A comparison between two empirical downscaling strategies", Int. J. Climatology, vol 21, Issue 13, pp.1645-1668. DOI 10.1002/joc.703]. and mixFields prepares for mixed-field EOF analysis [ref. Bretherton et al. (1992) "An Intercomparison of Methods for finding Coupled Patterns in Climate Data", J. Climate, vol 5, 541-560; Benestad et al. (2002), "Empirically downscaled temperature scenarios for Svalbard", Atm. Sci. Lett., doi.10.1006/asle.2002.0051].

Uncertainty estimates are computed according to North et al. (1982), "Sampling Errors in the Estimation of Empirical Orthogonal Functions", Mon. Weather Rev., vol 110, 699-706.

The EOFs are based on svd.

See the course notes from Environmental statistics for climate researchers http://www.gfi.uib.no/~nilsg/kurs/notes/course.html for a discussion on EOF analysis.

The method PCA is similar to EOF, but designed for parallel station series (e.g. grouped together with merge). PCA does not assume gridded values and hence does not weigth according to grid area. PCA is useful for downscaling where the spatial covariance/coherence is important, e.u involving different variables from same site, same variable from different sites, or a mix between these. For instance, PCA can be applied to the two wind components from a specific site and hence extract the most important wind directions/speeds.

Usage

EOF(X,it=NULL,n=20,lon=NULL,lat=NULL,verbose=FALSE,...)
EOF.default(X,it=NULL,n=20,lon=NULL,lat=NULL,
            area.mean.expl=TRUE,verbose=FALSE,...)
EOF.field(X,it=NULL,n=20,lon=NULL,lat=NULL,
          area.mean.expl=TRUE,verbose=FALSE)
EOF.comb(X,it=NULL,n=20,lon=NULL,lat=NULL,area.mean.expl=TRUE,verbose=FALSE)
eof2field(eof,is=NULL,lon=NULL,lat=NULL,anomaly=FALSE)
PCA(X,...) 
EOF.default(X,...)
PCA.station(X,na.action='fill',verbose=FALSE)
pca2station(X,lon=NULL,lat=NULL,anomaly=FALSE)

Arguments

X	a 'field' or 'pca' object
it	see subset
n	number of EOFs
is	Spatial subsetting - see subset.eof
lon	set longitude range - see t2m.ERAINT
lat	set latitude range
verbose	TRUE - clutter the screen with messages
area.mean.expl	When TRUE, subtract the area mean for the domain and use as a the first co-variate before the PCs from the EOF analysis.
na.action	'fill' uses approx to interpolate the NA-values before the PCA.

Value

File containing an 'eof' object which is based on the 'zoo' class.

Author(s)

R.E. Benestad

Examples

# Simple EOF for annual mean SST:
sst <- sst.NCEP(lon=c(-90,20),lat=c(0,70))
SST <- aggregate(sst, year, mean, na.rm = FALSE)
eof.sst <- EOF(SST)
plot(eof.sst)

# EOF of July SST:
eof.sst7 <- EOF(sst,it=7)
plot(eof.sst7)

# common EOF for model
# Get some sample data, extract regions:
GCM <- t2m.NorESM.M()
gcm <- subset(GCM,is=list(lon=c(-50,60),lat=c(30,70)))
t2m.eraint <- t2m.ERAINT()
eraint <- subset(t2m.eraint,is=list(lon=c(-50,60),lat=c(30,70)))

OBS <-  aggregate(eraint, by=year, mean, na.rm = FALSE)
GCM <- aggregate(gcm, by=year, mean, na.rm = FALSE)
OBSGCM <- combine(OBS,GCM,dimension='time')

ceof <- EOF(OBSGCM)
plot(ceof)

# Example for using PCA in downscaling
nacd <- station(src='nacd')
X <- annual(nacd)
nv <- function(x) sum(is.finite(x))
ok <- (1:dim(X)[2])[apply(X,2,nv) == dim(X)[1]]
X <- subset(X,is=ok)
pca <- PCA(X)
map(pca)

slp <- slp.NCEP(lon=c(-20,30),lat=c(30,70))
eof <- EOF(slp,it=1)
ds <- DS(pca,eof)
# ds is a PCA-object
plot(ds)

# Recover the station data:
Z <- pca2station(ds)
plot(Z)

metno/esd.test documentation built on May 22, 2019, 7:49 p.m.