yuming-jin / eof

Empirical Orthogonal Function to compute spatial and temporal pattern of geographic data

EOF

Empirical Orthogonal Function to compute spatial and temporal pattern of geographic data (Adapted from Wikle, C. K., Zammit-Mangion, A., and Cressie, N. (2019), Spatio-Temporal Statistics with R)

---

Required R packages

reshape2

---

Defination of Variables

stdata: 2D array of data, matrix format: space (row) x time (columun)

var1: value of first dimension of space (ex. longitude)

var2: value of second dimension of space (ex. latitude)

time: value of timestep

EOFnumber: number of EOF to obtain in the output

---

Example

library(ncdf4)

library(lubridate)

nc <- nc_open('uwnd.mon.mean.nc') #download from: https://psl.noaa.gov/data/gridded/data.ncep.reanalysis2.pressure.html

time <- ncvar_get(nc=nc,varid='time')

date <- as.POSIXct(as.character(as.Date(time/24,origin='1800-01-01')))

time <- decimal_date(as.POSIXct(as.character(as.Date(time/24,origin='1800-01-01'))))

lon <- ncvar_get(nc=nc,varid='lon')

lat <- ncvar_get(nc=nc,varid='lat')

p <- ncvar_get(nc=nc,varid='level')

u <- ncvar_get(nc=nc,varid='uwnd')

u <- u[,,which(p==500),]

nc <- nc_open('vwnd.mon.mean.nc') #download from: https://psl.noaa.gov/data/gridded/data.ncep.reanalysis2.pressure.html

v <- ncvar_get(nc=nc,varid='vwnd')

v <- v[,,which(p==500),]

wind <- u+1i*v

stdata <- u

var1 <- lon

var2 <- lat

time <- time

eof_output <- EOF(stdata = stdata,var1 = var1, var2 = var2, time = time,EOFnumber = 2)

EOFspace <- eof_output[[1]]

EOFtime <- eof_output[[2]]

EOFd <- eof_output[[3]]