[en] In the Czech Hydrometeorological Institute (CHMI) there exists a unique set of meteorological measurements consisting of the values of vertical atmospheric levels of beta and gamma radiation. In this paper a stochastic data-driven model based on nonlinear regression and on nonhomogeneous Poisson process is suggested. In the first part of the paper, growth curves were used to establish an appropriate nonlinear regression model. For comparison we considered a nonhomogeneous Poisson process with its intensity based on growth curves. In the second part both approaches were applied to the real data and compared. Computational aspects are briefly discussed as well. The primary goal of this paper is to present an improved understanding of the distribution of environmental radiation as obtained from the measurements of the vertical radioactivity profiles by the radioactivity sonde system. - Highlights: → We model vertical atmospheric levels of beta and gamma radiation. → We suggest appropriate nonlinear regression model based on growth curves. → We compare nonlinear regression modelling with Poisson process based modeling. → We apply both models to the real data.