EmpiricalOrthogonalFunctions.jl

Julia package for calculating Empirical Orthogonal Functions from spatiotemporal datasets.

This package was heavily inspired by the eofs Python package and a good amount of code was translated to Julia from this package.

Installation

using Pkg
Pkg.add("EmpiricalOrthogonalFunctions")

Example

This example will highlight extracting the spatial and temporal flooding signals from a series of satellite imagery over Southeast Asia

using EmpiricalOrthogonalFunctions
using NCDatasets

#load in the data
ds = NCDataset("sar_stack.nc","r");
datain = ds["VV"][:];

# apply EOF
eof = EmpiricalOrthogonalFunction(datain; timedim=3)

# rotate the EOFs using varimax rotations
nmodes = 4
reof = orthorotation(eof,n=nmodes)

# extract out the signals
# the spatial signals are reshaped back to the original dimensions
temporalsignal = pcs(reof)
spatialsignal = reshape(eofs(reof),(size(datain)[1:2]..., nmodes))

When plotting the first four spatial signals we will get the following plot.

Below is the temporal signals corresponding to the spatial signals above