CDO

Let’s demystify the cosmic dance of EOF (Empirical Orthogonal Function) modes and compute their variance percentages.

Eigenvalues and Variance:
Each EOF has an eigenvalue—a measure of its significance. Think of it as the groove factor in a cosmic dance-off!
To find the variance percentage contributed by an EOF, follow this celestial recipe:
Divide the eigenvalue of an EOF by the total variance of your data.
It’s like calculating the proportion of stardust that each dancer brings to the cosmic ball.
Total Variance Calculation:
You’ll need the total variance of your input data. Fear not! CDO (Climate Data Operators) can help:
Use cdo fldsum -timvar infile to calculate the total variance.
Set the environment variable CDO_WEIGHT_MODE=off to ensure correct results.
Variance Percentage Formula:
For each EOF (let’s say EOF1, EOF2, and EOF3):
Variance percentage of EOF1 = (Eigenvalue of EOF1) / (Total variance)
Variance percentage of EOF2 = (Eigenvalue of EOF2) / (Total variance)
Variance percentage of EOF3 = (Eigenvalue of EOF3) / (Total variance)
Example:
Suppose your EOF eigenvalues are:
Eigenvalue of EOF1 = 100
Eigenvalue of EOF2 = 50
Eigenvalue of EOF3 = 30
And the total variance is 500.
Then:
Variance percentage of EOF1 = (100 / 500) * 100 = 20%
Variance percentage of EOF2 = (50 / 500) * 100 = 10%
Variance percentage of EOF3 = (30 / 500) * 100 = 6%
Using CDO:
If you’re using CDO for EOFs, here’s a basic command snippet:
cdo eof,n input.nc eigval.nc pattern.nc
cdo eofcoeff pattern.nc input.nc pc.nc
cdo -div eigval.nc -timsum eigval.nc /explvar.nc

Replace n with the number of EOFs you want as output.
Remember:
The magnitude (and even the sign) of the EOF patterns matters.
Verify that your results align with your data and research context.