A total of 26 linear combinations (models) of location and scale parameters of GEV distribution were prepared using covariates ENSO, IOD, and AMO termed as C1, C2, and C3, respectively. Models formulated through various combinations of these covariates (M0, M1, M2, …) can be found in Table 3. Where M0 is the stationary model with constant parameters of GEV distribution; hence, independent of the LSCOs. However, total 26 nonstationary models (Table 3), starting from M1 to M26, in which scale and location parameters of GEV distribution have been represented as a function of covariates, are incorporated in this study. The shape parameter of the GEV distribution has been kept constant to avoid complexity in modelling (Coles 2001; Katz 2013; Yilmaz & Perera 2014; Das et al. 2020). The maximum-likelihood estimation (MLE) approach was used for parameter estimation from GEV distribution. Through MLE approach, the value of θ = [μ, σ, ξ] has been obtained at which maximum likelihood functions attain their maximum value. Let X = x₁, x₂, x₃, …., x(n − 1), x(n) be the series of any selected extreme with n number of observations. Then, the log-likelihood can be defined as (Katz 2013; Bracken et al. 2018):Here, L(θ) is the likelihood function of a particular parameter vector θ. The likelihood function is the measure of how likely the dataset represents as a function of unknown parameters of distribution (GEV, in the present case) and MLE gives the values of parameters that maximize the likelihood function (Katz 2013).