In this article, a modeling approach for the mean annual flow in different segments of Sabine river, as released in the NHDPlus data in 2007, as a function of five predictor variables is described. Modeling flow is extremely complex and the deterministic flow models are widely used for that purpose. The justification for using these deterministic models comes from the fact that the flow is governed by some explicitly stated physical laws. In contrast, in this article, this complex issue is addressed from a completely statistical point of view. A semiparametric model is proposed to analyze the spatial distribution of the mean annual flow of Sabine river. Semiparametric additive models allow explicit consideration of the linear and nonlinear relations with relevant explanatory variables. We use a conditionally specified Gaussian model for the estimation of the univariate conditional distributions of flow to incorporate auxiliary information and this formulation does not require the target variable to be independent.Research highlights► The complex issue of modeling flow is addressed from a completely statistical point of view. ► The conditionally specified Gaussian model is used to incorporate the dependence structure. ► A semiparametric additive model is used for the mean part in the conditionally specified Gaussian model.