“…Based on the assumption that S j is only related to the collocated H i = α ( j ) (Doyen et al., 1996), the secondary attribute data S j are assumed to be the summation of the vector of the model output and the noise vector : where f k is the mathematical model that relates the k th kind of secondary attribute to H i = α ( j ) , α is an operator used to transform j into the corresponding i at a given location, and is the misfit of the k th kind of secondary attribute data with its model output . Note that f k can be either a regression (J. Chen et al., 2001), geostatistical (Ruggeri et al., 2013), or physical inversion model (Zanini & Woodbury, 2016; Zanini et al., 2017). A regression model is utilized in this study.…”