“…Then, a data matrix of order n × p was developed, where n is the number of synoptic stations (139) and p (720) is monthly precipitation and temperature during a 30‐year period (1991–2020) (2 × 12 × 30 = 720). Based on the following equation, a typical PCA is applied to a p × p covariance (or the correlation) matrix to obtain the covariance matrix S (Bethere et al,
2017),
where X is the sample covariance matrix associated with the dataset and X T denotes its transpose. We can find eigenvectors ( e i , i = 1, …, 720) and corresponding eigenvalues ( λ i , i = 1, …, 720) by the following equation (Bethere et al,
2017):
where e is an eigenvector and λ is the corresponding eigenvalue of the covariance matrix S. We obtained non‐correlated linear combinations of the initial climatic variables using the following equation (Bethere et al,
2017):
where λ i represents the variance of each principal component Y i , and e i defines each PC called loadings.…”