EOF Analysis

This repository contains a Python implementation of Empirical Orthogonal Function (EOF) analysis, a widely used technique in climate and atmospheric sciences to extract dominant spatial patterns from geophysical datasets.

Features

• EOF Analysis: Perform Principal Component Analysis (PCA) on gridded data to identify modes of variability.
• Handles Latitude Weighting: Accounts for the cosine of latitude to apply proper area weighting.
• North Criterion: Implements the North et al. criterion for determining the significance of modes.
• Customizable: Allows for PCA to be performed on either the covariance matrix or the correlation matrix.

Requirements

Make sure you have the following Python packages installed:

• numpy
• scipy

You can install the dependencies by running:

pip install numpy scipy

Example

import numpy as np
from eof_analysis import perform_eof_analysis

# Example data: a 3D numpy array of shape [time, lat, lon]
# Replace this with your actual dataset
dat = np.random.rand(100, 20, 30)  # e.g., 100 time steps, 20 latitudes, 30 longitudes
lati = np.linspace(-90, 90, 20)  # 20 latitude points
loni = np.linspace(0, 360, 30)   # 30 longitude points

# Perform EOF analysis
score, latent, loading, north = perform_eof_analysis(dat, lati, loni)

# View results
print("Principal component time series (scores):", score)
print("Eigenvalues (latent):", latent)
print("EOF loading patterns:", loading)
print("North criterion results:", north)

Authors

• Husile Bai ([email protected])
• Chayan Roychoudhury ([email protected])