“…However, the reliability associated with BMA parameters has not received enough attention. In past literature relevant to the applications of BMA approach, both the weights and variances obtained through the Expectation-Maximization (EM) algorithm are assigned fixed values (Cao et al, 2021;Darbandsari and Coulibaly, 2021;Dormann et al, 2018;Duan et al, 2007;Huang and Merwade, 2023;Liu and Merwade, 2018;Madadgar and Moradkhani, 2014;Moknatian and Mukundan, 2023;Tsai, 2010). These fixed values cannot represent the uncertainty in the BMA parameters, especially when different datasets are used for training or a specific training dataset does encompass the overall prediction capacity of certain models (Madadgar and Moradkhani, 2014;Refsgaard et al, 2012;Rojas et al, 2010;Tebaldi and Knutti, 2007).…”